On Mon, Jun 24, 2024 at 4:46 PM Maxime Coquelin
wrote:
>
> This patch improves the FD manager logging in case
> of FD removal from the epoll FD set failure.
>
> When the FD is not more valid, like for example if it
is not valid anymore*
> has already been closed, only emit a debug log and not
>
> -Original Message-
> From: pbhagavat...@marvell.com
> Sent: Monday, June 24, 2024 2:53 PM
> To: Jerin Jacob ; Amit Prakash Shukla
>
> Cc: dev@dpdk.org; Pavan Nikhilesh Bhagavatula
>
> Subject: [PATCH] eventdev/dma: fix missing response info in forward mode
>
> From: Pavan Nikhilesh
>
On 6/24/24 17:02, Hernan Vargas wrote:
For perf tests, the operation size for the MLD-TS was incorrect.
Fixed so that the performance numbers are correct.
Largely cosmetic only.
Fixes: 95f192a40e35 ("test/bbdev: add MLD cases")
Cc: sta...@dpdk.org
Signed-off-by: Hernan Vargas
---
app/test
On 6/24/24 17:02, Hernan Vargas wrote:
Add check for FFT window width.
Signed-off-by: Hernan Vargas
---
app/test-bbdev/test_bbdev_perf.c | 26 ++
app/test-bbdev/test_bbdev_vector.c | 14 ++
app/test-bbdev/test_bbdev_vector.h | 2 ++
3 files changed
On Mon, Jun 24, 2024 at 6:54 PM Harman Kalra wrote:
>
> From: Hanumanth Pothula
>
> As part of ready and exit messages only first half of the ports
> were processed, i.e. represented to representor port mapping was
> setup and released. While later half of the ports were not
> processed.
>
> Fixe
The rte_vhost_driver_unregister() vhost_user_read_cb()
vhost_user_client_reconnect() can be called at the same time by 3 threads.
when memory of vsocket is freed in rte_vhost_driver_unregister(),
then vhost_user_read_cb() maybe add vsocket to reconn_list,
the invalid memory of vsocket is accessed i
This is a patchset loosely based on internal development tree which
tracked ICE driver development over a period of time since last
snapshot update.
v2-> v3:
- Fixed commit authorship for commits where it was feasible to do so
- Addressed review comments
- Clarified and corrected a few commit mess
From: Przemek Kitszel
Replace all usages of ICE_SUCCESS with its integer equivalent.
Signed-off-by: Przemek Kitszel
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_acl.c | 2 +-
drivers/net/ice/base/ice_acl_ctrl.c | 16 +--
drivers/net/ice/base/ice_common.c| 78 +++--
From: Paul Greenwalt
Add initial set of E830-related definitions.
Signed-off-by: Paul Greenwalt
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_hw_autogen.h | 79 +--
1 file changed, 50 insertions(+), 29 deletions(-)
diff --git a/drivers/net/ice/base/ice_hw_aut
From: Siddaraju DH
Update en_link in curr_user_phy_cfg.caps on successful execution of
set_link_restart_an operation.
Signed-off-by: Siddaraju DH
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_common.c | 12 +++-
1 file changed, 11 insertions(+), 1 deletion(-)
diff --git a/dr
From: Stefan Wegrzyn
Currently, the 'temp' variable is in the same scope as 'first_teid_ptr'
and we take point to it, which means on new loop iterations the memory
goes out of scope and we keep a dangling pointer. Move 'temp' to
outside of the loop to fix it.
Signed-off-by: Stefan Wegrzyn
Signe
From: Jesse Brandeburg
When building with certain targets, a missing include was causing a compilation
failure. Add the missing include.
Signed-off-by: Jesse Brandeburg
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_flow.h | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/ne
From: Zhirun Yan
Support to add large action to set 32 bits markid via switch filter.
Signed-off-by: Zhirun Yan
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_switch.c | 22 --
drivers/net/ice/base/ice_switch.h | 24
2 files changed, 24 ins
From: Jacob Keller
The ice_debug_cq function is called to print debug data for a control queue
descriptor in multiple places. This includes both before we send a message on a
transmit queue, after the writeback completion of a message on the transmit
queue, and when we receive a message on a rece
From: Fabio Pricoco
Add new Admin Q command to check if netlist auth is supported.
Signed-off-by: Fabio Pricoco
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_common.c | 5 +
drivers/net/ice/base/ice_type.h | 2 ++
2 files changed, 7 insertions(+)
diff --git a/drivers/net/ice/b
From: Fabio Pricoco
Initialize some variable to avoid "variable might be not defined" warning. This
is due to the fact that when using silicon validation tools, sometimes
alternative fragments of code are being executed exclusively thanks to use of
goto, which some compilers cannot handle correct
From: Paul Greenwalt
The use of get link status PHY type for determining the media type can result in
incorrect reporting the incorrect media type. This can happen for a couple
reasons:
1. The get link states current PHY type field can return non-PHY type values
(i.e. bit 127 indicates link sta
From: Dawid Zielinski
New flag for external topology devices for different types of versioning.
Signed-off-by: Dawid Zielinski
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_common.c | 6 ++
drivers/net/ice/base/ice_type.h | 2 ++
2 files changed, 8 insertions(+)
diff --git a/d
From: Ashish Shah
ice_flow_seg_info fields match and range are being used with bit
operations but not declared as bitmaps. This can cause issues when
casting values greater than 32. This change is to declare them as proper
bitmaps so that the bitmap operations can function as intended.
Signed-of
From: Jesse Brandeburg
The firmware interface definition file mistakenly started using bitmap
declarations for its storage in a couple of structures. The data structure
contains a "bit-map" of bits that the hardware/firmware sets or reads but it's
not a DECLARE_BITMAP style bitmap and it's not ac
From: Jacob Keller
The driver allocates a cmd_buf array in addition to the desc_buf array. This
array stores an ice_sq_cd command details structure for each entry in the
control queue ring.
The contents of the structure are copied from the value passed in via
ice_sq_send_cmd, and include only a
From: Michal Swiatkowski
When devlink reload is called port_info should stay. Check if it wasn't already
allocated before doing allocation.
Signed-off-by: Michal Swiatkowski
Signed-off-by: Anatoly Burakov
---
drivers/net/ice/base/ice_common.c | 3 ++-
1 file changed, 2 insertions(+), 1 deleti
From: Jesse Brandeburg
The Linux Kernel is now requiring flex array safe allocations from drivers,
so the code is being updated to reflect the changes required to match Linux
upstream code.
Signed-off-by: Jesse Brandeburg
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_adminq_cmd.h |
From: Fabio Pricoco
OROM update in recovery mode is allowed or not based on function capability
0x0090. This patch adds capability parsing for it.
Signed-off-by: Fabio Pricoco
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_adminq_cmd.h | 1 +
drivers/net/ice/base/ice_common.c | 5
From: Grzegorz Nitka
Return error if there were no recipes found.
Signed-off-by: Grzegorz Nitka
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_switch.c | 5 +
1 file changed, 5 insertions(+)
diff --git a/drivers/net/ice/base/ice_switch.c
b/drivers/net/ice/base/ice_switch.c
index
From: Eric Joyner
If ice_cfg_phy_fec() goes through case ICE_FEC_DIS_AUTO but the underlying
device firmware isn't the correct version, then the memory used to hold the PHY
capabilities output from the firmware isn't freed.
Fix this by making sure the memory is freed by jumping to the label that
From: Karol Kolacinski
E810 products can support low latency Tx timestamp register read using the SW
interrupt from the FW. Add a check for the device capability and use the new
method if supported.
Signed-off-by: Karol Kolacinski
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_common.
From: Sergey Temerkhanov
Set up and use model-dependent number of PHY ports.
Also fix iterations over PTP ports.
Define 'max_phy_port', a new ice_hw struct member to store the max port number
handled by given clock owner. Replace 'phy_ports' with 'max_phy_port' for max
range in iterations over
From: Mateusz Pacuszka
Currently all functions related to promiscuous modes use u8 bit mask. All 8 bits
are utilized already and to be able to add any new flag, all the drivers need to
comply.
To address that, change u8 into ice_bitmap_t and use dedicated functions to
process it.
Signed-off-by:
From: Mateusz Pacuszka
Currently for unicast promiscuous, the driver is adding a LOOKUP_RX filter rule
using recipe 3. The direction bit in this recipe is the pkt_is_from_network
flag, so the filter only matches on packets from the network, not loopback
packets. To match loopback packets without
From: Michal Michalik
E825-C hardware has a new feature which allows to change the clocks which drives
the TX on the PHYs. The registers are accessed directly no interaction with FW
via admin queue commands. Add this support.
Signed-off-by: Michal Michalik
Signed-off-by: Ian Stokes
---
driver
From: Robert Malz
When applying UDP cloud filter with tunnel type different than VXLAN for each of
the filter new recipe will be created instead of using already created one. This
will quickly use all of the resources for specific recipe. Modified check to
detect if recipe with similar tunnel typ
From: Lukasz Plachno
The original bit-shift was made on a 16-bit value which could overflow.
Use 64-bit integer to do bit-shifts instead.
Signed-off-by: Lukasz Plachno
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_parser_rt.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
From: Michal Michalik
The new series of devices supported by ice driver are E830 devices. This
hardware supports the PTP feature very similar to E810 devices. The main
difference besides some other register addresses is that in E830 we are able to
read the TX timestamps memory directly from BARs
From: Fabio Pricoco
When errors happen during OROM reads, the allocated buffer is not freed.
Fix it by adding a goto label to free the buffer in case of error.
Signed-off-by: Fabio Pricoco
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_nvm.c | 10 ++
1 file changed, 6 insertio
From: Jesse Brandeburg
When previous version of driver snapshot was generated, there was a "fix"
applied to it that moved the lock inside the if-else block. This change was made
in error and was later reverted in the base driver, so adjust accordingly.
Signed-off-by: Jesse Brandeburg
Signed-off
From: Jacob Keller
Since we know that most side band queue messages complete within 2-3
microseconds, introduce an initial 5 microsecond delay before we enter the main
timeout loop. Use ice_flush(hw) first to ensure that we immediately flush the
tail bump before delaying. This should mean that in
From: Stefan Wegrzyn
Adjust ice_aq_get_internal_data() wrapping function so that it returns next
cluster ID when iterating over clusters, as well as add missing cluster
definition.
Signed-off-by: Stefan Wegrzyn
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_adminq_cmd.h | 4 ++--
dri
From: Karol Kolacinski
With the introduction of PTP auxiliary bus driver, aqc_driver_params are no
longer needed to share data between the PFs.
Signed-off-by: Karol Kolacinski
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_adminq_cmd.h | 5 -
1 file changed, 5 deletions(-)
diff -
From: Yochai Hagvi
Data should not be changed by generic i2c write, hence it should be const.
Signed-off-by: Yochai Hagvi
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_common.c | 2 +-
drivers/net/ice/base/ice_common.h | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --
From: Jesse Brandeburg
Fix a static analysis warning where if the 16-bit value in mask has the high-bit
set, it will be sign extended by the shift left (which converts it to a signed
integer). Avoid this by casting to a u32 to make sure the conversion happens
before the shift and that it stays un
From: Yochai Hagvi
Add a helper function to determine whether a given pin of DPLL is capable for
refsync configuration.
Signed-off-by: Yochai Hagvi
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_ptp_hw.c | 16
drivers/net/ice/base/ice_ptp_hw.h | 1 +
2 files changed,
From: Prathisna Padmasanan
Signed-off-by: Prathisna Padmasanan
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_common.c | 4
1 file changed, 4 insertions(+)
diff --git a/drivers/net/ice/base/ice_common.c
b/drivers/net/ice/base/ice_common.c
index ff98919e40..61cf6da3ad 100644
---
From: Wojciech Drewek
Define Cage Max Power override NVM module which allows to override maximum power
of QSFP/SFP cage. There might be up to 8 cages so each of them has its own word
in the module. From the ice driver perspective we need only Override Enable bit
so there is no need to define bits
From: Eric Joyner
Previously, the "No FEC in Auto mode" feature was added to the link management
engine (part of firmware) for E810 devices, and was unsupported at the time on
E82X devices. However, this support is now being extended to new FW versions for
E82X devices, and so the SW check to ena
From: Jacob Keller
The ice_sched_add_root_node() and ice_sched_add_node() functions have comments
to suppress Coverity warnings about a suspicious sizeof used when allocating the
children array of an struct ice_sched_node.
The size is calculated using the size of the scheduler node, which overal
From: Eric Joyner
Update the offsets, and remove the ICE_TX_GCS_DESC_ENA define since a non-0
value used for the ICE_TX_GCS_DESC_TYPE field will enable GCS offload; there is
no dedicated bit to enable it anymore.
Signed-off-by: Eric Joyner
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ic
From: Piotr Raczynski
Add a separate VSI type for devlink subfunctions due to specific handling in the
driver.
Signed-off-by: Piotr Raczynski
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_type.h | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/net/ice/base/ice_type.h b/dri
From: Chinh Cao
"count" is the returned value, and in this function, value of "count" and "bit"
are the same, so there is no reason why these values should be declared as
different types, and for returned value data type to be different.
Signed-off-by: Chinh Cao
Signed-off-by: Ian Stokes
---
From: Jacob Keller
Code using ice_ptp_one_port_cmd() (or the device specific variants for E822 and
ETH56G) has a subtle bug where-in the executing of ice_ptp_exec_tmr_cmd() will
cause all other ports to execute their last executed command.
Most flows operate on all timers at once, but some flows
From: Jacob Keller
A couple of E822 PTP functions use ICE_PTP_READ_TIME on the main timer to avoid
any side effects when programming a single PHY port. Match the ETH56G code by
using ICE_PTP_NOP instead, which better reflects the intended behavior.
Signed-off-by: Jacob Keller
Signed-off-by: Ian
From: Adrian Pielech
Removed "else" statement in ice_get_ddp_pkg_state() as it may never happen.
Signed-off-by: Adrian Pielech
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_ddp.c | 5 +
1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/drivers/net/ice/base/ice_ddp.c b/d
From: Jacob Keller
With the exception of a few functions, the upstream Linux version of
ice_ptp_hw.c switched to using "int err" early before the initial conversion of
"enum ice_status status" to "int status" was completed. This results in the
upstream version using a different variable name than
From: Fabio Pricoco
Shared code has a fw_log.c file that is used to handle the firmware
logging. For DPDK it's empty, but it is added to reduce delta between
DPDK and shared code.
Signed-off-by: Fabio Pricoco
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_fwlog.c
From: Jacob Keller
When upstreaming the E810-T support for Linux, during review it was pointed out
that the E810T subdevice IDs numbering was odd. The subdevice ID numbers seemed
to be arbitrary. The upstream driver re-numbered the subdevice ID values so that
they were in increasing order.
Align
From: Jacob Keller
In certain unload conditions the non-standard paths to breaking down the bond
supporting resources need special considerations in that there could be nodes
moved out of the tree and this can cause a NULL pointer to be passed into the
find node by
teid function.
Add in a NULL p
From: "Temerkhanov, Sergey"
Check for 1PPS GPIO association in 1588 function caps. Remove legacy code
related to a workaround setting 1588 capabilities to a fixed value.
Signed-off-by: Temerkhanov, Sergey
Signed-off-by: Nitka, Grzegorz
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_c
From: Karol Kolacinski
If TS PLL fails to lock to the values from the capabilities, retry with safe
internal 25 MHz TCXO in nanosecond mode. This prevents PTP init fail and does
not disturb nominal initialization of the device.
Signed-off-by: Karol Kolacinski
Signed-off-by: Ian Stokes
---
dri
From: Mateusz Pacuszka
In case the rule already exists and another VSI wants to subscribe to it new VSI
list is being created and both VSIs are moved to it. Currently, the check for
already existing VSI with the same rule is done based on fdw_id.hw_vsi_id, which
applies only to LOOKUP_RX flag. Ch
From: Yochai Hagvi
When converting to source timer command value, tmr_idx is left shifted by 8
(CPK_SRC_SEL), which overflows the 8-bit data type. Change to 32-bit.
Signed-off-by: Yochai Hagvi
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_ptp_hw.c | 3 +--
1 file changed, 1 insertion
From: Dave Ertman
In the BW configuration performed by DCF functions, the strict/WFQ and priority
field (referred to as Generic in the EAS) is not updated in the FW. This needs
to be updated so as to not incorrectly allocate BW credits in the traffic
shaping Tx scheduler.
Signed-off-by: Dave Er
From: Bartosz Staszewski
Previous implementation of PHY timestamp retry value was set to a static value
that had no meaning. Change it to calculate it based on a set of meaningfully
named macros, as well as adjust data type to avoid overflows.
Signed-off-by: Bartosz Staszewski
Signed-off-by: Ia
From: NorbertX Ciosek
Check if recipe buffer was already allocated before allocating new one.
Without such check it is possible to get memory leak when overwriting
pointer address.
Signed-off-by: NorbertX Ciosek
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_switch.c | 3 +++
1 file c
From: Fabio Pricoco
num_rules was defined u16 but it should be u8 as it is passed to
function ice_aq_sw_rules which expects u8.
Signed-off-by: Fabio Pricoco
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_switch.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/driv
From: Waldemar Dworakowski
Some compilation targets were generating errors due to unchecked array bounds
when accessing recipe groups. Add a check to avoid that.
Signed-off-by: Waldemar Dworakowski
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_switch.c | 14 --
1 file cha
From: Jacob Keller
Add an informative error message when preparing external PHYs for a timer
sync fails.
Signed-off-by: Jacob Keller
Signed-off-by: Anatoly Burakov
---
drivers/net/ice/base/ice_ptp_hw.c | 11 +--
1 file changed, 9 insertions(+), 2 deletions(-)
diff --git a/drivers/net
From: Dawid Zielinski
Coppied output IO params returned from get netlist node pin command descriptor
so SW could use it. Fixed defines related to output IO params.
Signed-off-by: Dawid Zielinski
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_adminq_cmd.h | 4 ++--
1 file changed, 2 in
From: Waldemar Dworakowski
Certain compilation targets were generating compile errors due to snprintf
return value not being checked. Fix it by ignoring the return value explicitly.
Signed-off-by: Waldemar Dworakowski
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_common.c | 2 +-
1 f
From: Jaroslaw Ilgiewicz
Previously NVM was once acquired and OROM was read to RAM in 4KB chunks.
Changed this to a loop in which with each iteration NVM is acquired and
only 4 bytes are read.
Signed-off-by: Jaroslaw Ilgiewicz
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_nvm.c | 43
From: Paul Greenwalt
It's possible that an NVM with an invalid tlv_len could cause an integer
overflow of next_tlv which can result an infinite loop.
Fix this issue by changing next_tlv from u16 to u32 to prevent overflow.
Also check that tlv_len is valid and less than pfa_len.
Fix an issue wit
From: Remigiusz Konca
Updated data sheet added a new capability 'Next Cluster ID' defined as 0x0096,
so add a define for it.
Signed-off-by: Remigiusz Konca
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_adminq_cmd.h | 2 +-
drivers/net/ice/base/ice_common.c | 5 +
drivers/net/
From: Tomaszx Wakula
Add API and definitions related to reading Software Definable Pin section from
NVM, related to PTP pins assignment.
Signed-off-by: Tomaszx Wakula
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_adminq_cmd.h | 23 +++
drivers/net/ice/base/ice_ptp_hw.c | 87 +
From: Pabitra Dalai
To reflect updates in the data sheet, add:
- loopback mode value into ice_aqc_set_port_params
- required defines to support loopback mode
- added loopback mode handling in ice_port_info
Signed-off-by: Pabitra Dalai
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_ad
From: Pawel Sobczyk
Added sw_id variables to ice_fltr_info struct and disable lan_en flag for
Floating VEB VSIs.
Signed-off-by: Pawel Sobczyk
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_switch.c | 6 --
drivers/net/ice/base/ice_switch.h | 3 +++
2 files changed, 7 insertions(+)
From: Stefan Wegrzyn
As per updated data sheet, add 'skip_clear_pf' field to ice_hw structure, which
can be used to skip call to ice_clear_pf_cfg() in ice_init_hw(). Also, make
'fw_vsi_num' field of ice_hw structure visible to every component using shared
code.
Signed-off-by: Stefan Wegrzyn
Sig
From: Stefan Wegrzyn
Driver tools need to call ice_clear_vsi_q_ctx() from outside of ice_switch.c so
the function had to become non-static.
Signed-off-by: Stefan Wegrzyn
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_switch.c | 2 +-
drivers/net/ice/base/ice_switch.h | 1 +
2 files ch
From: Waldemar Dworakowski
The new version is compliant with implementation and documentation
Signed-off-by: Waldemar Dworakowski
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_adminq_cmd.h | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ice/base/ic
From: Dan Nowlin
Previously, the code would assume that only "Modular Signature Segment"
existed. Given a package with both a "Reference Signature Segment" and a
"Modular Signature Segment" the download would not have been successful
because an incorrect sequence of buffers would be sent to the f
From: Dan Nowlin
As a temporary workaround to allow unsigned packages, there
needs to be support for configuration segments that are unsigned.
Allow packages that have some configuration segments that are signed
and some configuration segments that are not signed. The code has
been changed to de
From: Stefan Wegrzyn
Wrong type specifiers were used in ice_write_*(), ice_read_*() and
ice_sched_add_root_node() functions. When using silicon validation tools, those
functions are operating on user space buffers and therefore the ice_memcpy type
has to be set to ICE_NONDMA_TO_NONDMA.
Signed-of
From: Dan Nowlin
Allow the driver to be compatible with different FW API versions based on the
device's MAC type. Currently, E810 is only compatible with one FW API version.
Now the driver can be compatible with different FW API versions for both E810
and E830.
Signed-off-by: Dan Nowlin
Signed-
From: Dan Nowlin
Add ice_pcie_speed_32_0GT bus speed enum to support E830 device's higher
bus speed.
Signed-off-by: Dan Nowlin
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_type.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/net/ice/base/ice_type.h b/
From: Ian Stokes
Add E830 debug dump cluster ID values, and rename E810 cluster IDs to make it
easier to distinguish which values are which.
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_adminq_cmd.h | 37 ++-
drivers/net/ice/ice_diagnose.c| 2 +-
2 fi
From: Fabio Pricoco
Add support for AQ 0x070C. Includes ice_nvm_sanitize and a helper
function ice_nvm_sanitize_operate to facilitate the operate request.
Signed-off-by: Fabio Pricoco
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_adminq_cmd.h | 25
drivers/net/ice/base/i
From: Tomaszx Wakula
Preparing a port for timesync command procedure shall provide Rx type which
indicates the type for the Rx metadata in the ONPI interface. According to the
documentation, Rx type should always equal 0x18.
Signed-off-by: Tomaszx Wakula
Signed-off-by: Ian Stokes
---
drivers/
From: Dan Nowlin
Add the 200G speeds to the ice_get_link_speed_based_on_phy_type function. This
allows detection of 200G speed instead of UKNOWN.
Signed-off-by: Dan Nowlin
Signed-off-by: Alice Michael
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_adminq_cmd.h | 1 +
drivers/net/ice
From: Karol Kolacinski
Add a possibility to mark all transmitted/received timestamps as invalid
by clearing PHY OFFSET_READY registers.
Signed-off-by: Karol Kolacinski
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_ptp_hw.c | 84 +++
drivers/net/ice/base/ic
From: Sergey Temerkhanov
Rename the PHY model designating fields, functions, etc to make the code more
readable and distinguish it from the "PHY type" as described in the
documentation.
Signed-off-by: Sergey Temerkhanov
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_ptp_hw.c | 44
From: Sergey Temerkhanov
Explicitly enable sideband device access before the first PHY access to avoid
possible errors for ETH56G.
Signed-off-by: Sergey Temerkhanov
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_ptp_hw.c | 38 +++
1 file changed, 29 inserti
From: Sergey Temerkhanov
Assign the ETH56G PHY address according to the PCI IDs in preparation for
further changes.
Signed-off-by: Sergey Temerkhanov
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_ptp_hw.c | 39 +++---
drivers/net/ice/base/ice_ptp_hw.h | 2 +
From: Sergey Temerkhanov
Support configurations with multiple ETH56G PHYs per a MAC complex. These PHYs
have their own addresses and are supposed to have their ports numbered
contiguously. This commit mainly provides port number translation for these
configurations.
Also, handle dual NAC correct
From: Karol Kolacinski
Add an interface to clear the timestamp memory block for E822 and ETH56G
based devices.
Signed-off-by: Karol Kolacinski
Signed-off-by: Ian Stokes
---
drivers/net/ice/base/ice_ptp_hw.c | 65 +++
drivers/net/ice/base/ice_ptp_hw.h | 2 +
2 file
From: Jacob Keller
The eth56g register checks for 40b and 64b registers do not return the high
address offset. Instead it is expected that all offsets are exactly 4 bytes
apart in the expected order. The e822 implementations explicitly return the
address in the lookup function. This form is bette
From: Jacob Keller
Both e822 and eth56g based hardware have a register indicating the status
of timestamps in the Tx timestamp memory bank. This register is used to
indicate which timestamps are currently valid. This is important because
the interrupt logic for these devices assumes that software
From: Sergey Temerkhanov
Implement upper-level PHY control functions as well as required device specific
defines. These include functions returning basic 56G PHY timing parameters and
definitions for constant parameters.
Signed-off-by: Sergey Temerkhanov
Signed-off-by: Ian Stokes
---
drivers/
From: Jacob Keller
This commit has multiple fixes related to comments updates, possible unitialized
variable use and finally a bug fix related to stale PHY commands. Details of the
bug fix are included below as it represents the bulk of the changes.
Code using ice_ptp_one_port_cmd() (or the devi
From: Aviraj CJ
Currently, sending a command to Sideband Queue (SBQ) is done with
unconditionally setting `ICE_AQ_FLAG_RD` flag. This patch allows the caller to
pass flags to the SBQ command function, and amends all call sites as needed.
Signed-off-by: Aviraj CJ
Signed-off-by: Ian Stokes
---
Upstream code has this function in a different place, so move it to
match the upstream code.
Signed-off-by: Anatoly Burakov
---
drivers/net/ice/base/ice_ptp_hw.c | 127 +++---
1 file changed, 63 insertions(+), 64 deletions(-)
diff --git a/drivers/net/ice/base/ice_ptp_hw.
From: Paul Greenwalt
E830 introduces a new version of Get Link Status Data which increases
the size of struct ice_aqc_get_link_status_data from 32 bytes to 51
bytes.
Add new link partner fields to ice_aqc_get_link_status_data; PHY type,
FEC, and flow control. Add new ICE_GET_LINK_STATUS_DATA_V2
Adds the Set/Get Port Options commands (0x06EA/0x06EB), also include
additional status and ID bits to improve the Link topology configuration
as well as ability to change port option from pre-boot environment via
the UEFI driver.
Signed-off-by: Anatoly Burakov
---
drivers/net/ice/base/ice_adminq
1 - 100 of 198 matches
Mail list logo
