Backing store type count can increase with newer firmware.
So a static bnxt_backing_store_types can cause unnecessary faults.
Instead of logging the string for the backing store type, just log
the value provided by the firmware.
Fixes: fe2f715ca580 ("net/bnxt: support backing store v2")
Cc: sta...
From: Damodharam Ammepalli
During bnxt_hwrm_func_backing_store_types_count driver
increments types only for valid type but not in
bnxt_hwrm_func_backing_store_cfg_v2. Implement the same
logic in store cfg also to prevent ctx types control loop
exiting the loop prematurely while firmware still has
Fix vector code to update the producer index correctly especially
for P7 devices where the code needs to handle epoch bit indication
correctly to the hardware.
Fixes: 30656a1cace8 ("net/bnxt: refactor epoch setting")
Cc: sta...@dpdk.org
Signed-off-by: Ajit Khaparde
---
drivers/net/bnxt/bnxt_rin
Some bug fixes for the bnxt PMD.
Please apply.
Ajit Khaparde (2):
net/bnxt: fix vector mode for P7 devices
net/bnxt: fix backing store logging
Damodharam Ammepalli (1):
net/bnxt: refactor backing store qcaps v2
drivers/net/bnxt/bnxt_hwrm.c| 64 +++--
driver
Add a new TestPmdPortStats data structure to represent the output
returned by `show port stats`, which is implemented as part of
TestPmdShell.
Bugzilla ID: 1407
Signed-off-by: Luca Vizzarro
Reviewed-by: Paul Szczepanek
---
dts/framework/remote_session/testpmd_shell.py | 75 +++
Add a new TestPmdPort data structure to represent the output
returned by `show port info`, which is implemented as part of
TestPmdShell.
The TestPmdPort data structure and its derived classes are modelled
based on the relevant testpmd source code.
This implementation makes extensive use of regula
The first line of the InteractiveShell send_command method is generally
the command input field. This sometimes is unwanted, therefore this
commit enables the possibility of omitting the first line from the
returned output.
Signed-off-by: Luca Vizzarro
Reviewed-by: Paul Szczepanek
---
dts/frame
Adds parsing text into a custom dataclass. It provides a new
`TextParser` dataclass to be inherited. This implements the `parse`
method, which combined with the parser functions, it can automatically
parse the value for each field.
This new utility will facilitate and simplify the parsing of compl
v5:
- fixed docstrings typos
v4:
- fixed up TextParser class wording
v3:
- fixed docstrings
- changed TextParser.compose signature
- explained TextParser further
- introduced new DTSError InternalError
v2:
- refactored parsing utility
- changed functionality of the parser for conciseness
- added a
When sending a command using an instance of InteractiveShell the output
should filter out the trailing shell prompt when returning it. After
every command two shell prompts are summoned. One is consumed as it is
used as a delimiter for the command output. The second one is not
consumed and left for
On 06/06/2024 19:52, Jeremy Spewock wrote:
+@classmethod
+def from_str(cls, text: str):
+match text:
+case "black":
+return cls.BLACK
+case "white":
+
On Thu, Jun 6, 2024 at 5:17 AM Luca Vizzarro wrote:
>
> Add a new TestPmdPortStats data structure to represent the output
> returned by `show port stats`, which is implemented as part of
> TestPmdShell.
>
> Bugzilla ID: 1407
>
> Signed-off-by: Luca Vizzarro
> Reviewed-by: Paul Szczepanek
Review
On Thu, Jun 6, 2024 at 5:17 AM Luca Vizzarro wrote:
>
> Add a new TestPmdPort data structure to represent the output
> returned by `show port info`, which is implemented as part of
> TestPmdShell.
>
> The TestPmdPort data structure and its derived classes are modelled
> based on the relevant testp
Just a few very minor documentation catches, otherwise:
Reviewed-by: Jeremy Spewock
On Thu, Jun 6, 2024 at 5:17 AM Luca Vizzarro wrote:
> +@classmethod
> +def from_str(cls, text: str):
> +match text:
> +case "black":
>
On Thu, Jun 6, 2024 at 5:17 AM Luca Vizzarro wrote:
>
> The first line of the InteractiveShell send_command method is generally
> the command input field. This sometimes is unwanted, therefore this
> commit enables the possibility of omitting the first line from the
> returned output.
>
> Signed-o
On Thu, Jun 6, 2024 at 5:17 AM Luca Vizzarro wrote:
>
> When sending a command using an instance of InteractiveShell the output
> should filter out the trailing shell prompt when returning it. After
> every command two shell prompts are summoned. One is consumed as it is
> used as a delimiter for
On 30. 5. 2024 17:25, Luca Vizzarro wrote:
Interactive shells that inherit DPDKShell initialise their params
classes from a kwargs dict. Therefore, static type checking is
disabled. This change uses the functionality of Unpack added in
PEP 692 to re-enable it. The disadvantage is that this fun
diff --git a/dts/framework/remote_session/dpdk_shell.py
b/dts/framework/remote_session/dpdk_shell.py
new file mode 100644
index 00..25e3df4eaa
--- /dev/null
+++ b/dts/framework/remote_session/dpdk_shell.py
@@ -0,0 +1,104 @@
+# SPDX-License-Identifier: BSD-3-Clause
+# Copyright(c) 2024
On 30. 5. 2024 17:25, Luca Vizzarro wrote:
Update the buffer scatter test suite to use TestPmdParameters
instead of the StrParams implementation.
Signed-off-by: Luca Vizzarro
Reviewed-by: Paul Szczepanek
Reviewed-by: Juraj Linkeš
Apparently I was replying to v2 so fixing that from now on.
On 30. 5. 2024 17:25, Luca Vizzarro wrote:
Implement all the testpmd shell parameters into a data structure.
Signed-off-by: Luca Vizzarro
Reviewed-by: Paul Szczepanek
Just one minor correction, otherwise:
Reviewed-by: Juraj Linkeš
For session-less crypto operations, event info is contained in
crypto op metadata for each event which is restored in event
from the crypto op metadata response info.
For session based crypto operations, crypto op contains per session
based event info in crypto op metadata. If any PMD passes any
i
Hi Abhi,
> -Original Message-
> From: Gujjar, Abhinandan S
> Sent: Thursday, June 6, 2024 5:47 PM
> To: Kundapura, Ganapati ; dev@dpdk.org;
> jer...@marvell.com
> Cc: Jayatheerthan, Jay ; Naga Harish K, S V
>
> Subject: RE: [PATCH v3] eventdev/crypto: restore opaque field between
> deque
On Wed, Jun 5, 2024 at 9:55 AM Patrick Robb wrote:
>
>
>
> On Fri, May 31, 2024 at 12:44 PM Luca Vizzarro wrote:
>>
>>
>> In my testing of Jeremy's patches which depend on this one ("Add second
>> scatter test case"), I've discovered that the Intel E810-C NIC I am
>> using to test does not automa
For session-less crypto operations, event info is contained in
crypto op metadata for each event which is restored in event
from the crypto op metadata response info.
For session based crypto operations, crypto op contains per session
based event info in crypto op metadata. If any PMD passes any
i
> >> I have added macros to help find the parameters and I have added mempool
> >> functions that allow you to determine if you can use the mempool and
> >> what params it needs. The new mempool functions are mentioned in the
> >> docs for ptr compress.
> >> Please take a look at v11.
> >
> > Gre
From: Shai Brandes
Replaced `enable_llq`, `normal_llq_hdr` and `large_llq_hdr`
devargs with a new shared devarg named `llq_policy` that
implements the same logic and accepts the following values:
0 - Disable LLQ.
Use with extreme caution as it leads to a huge performance
degradation on AW
From: Shai Brandes
Hi everyone, this is an API change to the ena PMD devargs that removes
the 3 devargs that control the user setting for the LLQ policy
and replaces it with a new shared devarg that implements the same logic.
Shai Brandes (1):
net/ena: restructure the llq policy user setting
On Thu, Jun 6, 2024 at 2:27 PM David Marchand wrote:
>
> When starting two DPDK programs using the same DPDK prefix (like for
> example OVS and testpmd, both running as primary processes in
> --in-memory mode), the first DPDK process of the two spews some error
> log when the second starts:
>
> TE
> +/**
> + * Compress pointers into 32-bit offsets from base pointer.
> + *
> + * @note It is programmer's responsibility to ensure the resulting offsets
> fit
> + * into 32 bits. Alignment of the structures pointed to by the pointers
> allows
> + * us to drop bits from the offsets. This is co
On 9. 5. 2024 13:20, Luca Vizzarro wrote:
Remove the imports in the testbed_model and remote_session modules init
file, to avoid the initialisation of unneeded modules, thus removing or
limiting the risk of circular dependencies.
Signed-off-by: Luca Vizzarro
Reviewed-by: Paul Szczepanek
---
On 9. 5. 2024 13:20, Luca Vizzarro wrote:
Move EalParams to its own module to avoid circular dependencies.
Maybe the commit message could mention that we added defaults.
Signed-off-by: Luca Vizzarro
Reviewed-by: Paul Szczepanek
Other than that,
Reviewed-by: Juraj Linkeš
Release status meeting minutes 2024-06-06
=
Agenda:
* Release Dates
* Subtrees
* Roadmaps
* LTS
* Defects
* Opens
Participants:
* AMD
* ARM
* Canonical
* Intel
* Marvell
* Nvidia
* Red Hat
Release Dates
-
The following are the current/updated
On 9. 5. 2024 13:20, Luca Vizzarro wrote:
Make it so that interactive shells accept an implementation of `Params`
for app arguments. Convert EalParameters to use `Params` instead.
String command line parameters can still be supplied by using the
`Params.from_str()` method.
Signed-off-by: Luca
Change flow_drv_list_create/destroy to mlx5_flow_list_create/destroy.
Remove resource release function inlining.
Check number of queues in template mode in hw configure function.
Use user priority to calculate matcher priority.
Signed-off-by: Maayan Kashani
Acked-by: Dariusz Sosnowski
---
driv
For ASO max allocations in non-template mode,
Read FW capabilities instead of using consts.
Signed-off-by: Maayan Kashani
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx5.h| 17 -
drivers/net/mlx5/mlx5_flow_hw.c| 13 +
drivers/net/mlx5/mlx5_flow_
Add support for encap/decap/modify header
action for non template API.
Save 1 action per bulk according to action data.
Reuse action if possible.
Store actions same as for SWS today, use same key structure.
Signed-off-by: Maayan Kashani
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx5_flo
Need to read maximum connection tracking
objects from HCA attributes.
Signed-off-by: Maayan Kashani
Acked-by: Dariusz Sosnowski
---
drivers/common/mlx5/mlx5_devx_cmds.c | 3 +++
drivers/common/mlx5/mlx5_devx_cmds.h | 1 +
2 files changed, 4 insertions(+)
diff --git a/drivers/common/mlx5/mlx5_d
reuse HWS actions translation code in order to create
the actions for lower layer.
All actions are handled as non masked.
Therefore, will be translated in the rule construct
phase shared with the template code
and not during the action template translate.
Currently there is no option to reconfigur
If the counter pool was not added to list,
and an error state was reached,
on attempt to destroy the counter pool,
segmentation fault was received during list remove action.
Added a check to verify the list is not empty before trying to
remove the cpool from the list.
Invalid state, leading to se
This patch adds counter, connection tracking, meter
and age actions support to non-template mode.
For CT, counter and meter: if no previous allocation was
handled by hw configure routine,
Half of the maximum supported number of objects will be allocated.
For AGE action, if no counters were allocat
From: Bing Zhao
The validation will be handled later.
Now it only supports the SEND_TO_KERNEL on non-root tables.
Signed-off-by: Bing Zhao
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx5_flow_hw.c | 23 ---
1 file changed, 20 insertions(+), 3 deletions(-)
diff --gi
From: Bing Zhao
Support JUMP / DROP / QUEUE / MARK / FLAG now.
Signed-off-by: Bing Zhao
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx5_flow_hw.c | 153
1 file changed, 135 insertions(+), 18 deletions(-)
diff --git a/drivers/net/mlx5/mlx5_flow_hw.c b/dr
From: Bing Zhao
The RTE_FLOW_ACTION_TYPE_END action needs to be translated into the
MLX5DR_ACTION_TYP_LAST for the future usage. It is only needed in
the hardware steering backward compatible API.
Signed-off-by: Bing Zhao
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx5.h | 1 +
Implement the frame and needed building
blocks for non template to hws API's.
Added validate, list_create and list_destroy to mlx5_flow_hw_drv_ops.
Rename old list_create/list_destroy functions to legacy_*
and added a call from verbs/dv ops to the legacy functions.
Updated rte_flow_hw as needed.
> Add two functions:
> - rte_mempool_get_mem_range - get virtual memory range
> of the objects in the mempool,
> - rte_mempool_get_obj_alignment - get alignment of
> objects in the mempool.
>
> Add two tests that test these new functions.
LGTM in general, few nits/suggestions below.
>
> Sign
When starting two DPDK programs using the same DPDK prefix (like for
example OVS and testpmd, both running as primary processes in
--in-memory mode), the first DPDK process of the two spews some error
log when the second starts:
TELEMETRY: Socket write base info to client failed
This is because t
Hi Ganapati,
Few comments inline. Please update in the next patch and add Acked-by:
Abhinandan Gujjar
> -Original Message-
> From: Kundapura, Ganapati
> Sent: Thursday, June 6, 2024 5:17 PM
> To: dev@dpdk.org; jer...@marvell.com; Gujjar, Abhinandan S
>
> Cc: Jayatheerthan, Jay ; Naga
>> +Prerequisites
>> +-
>> +
>> +This PMD driver need NPSDK library for system initialization and allocation
>> of resources.
>> +Communication between PMD and kernel modules is mediated by zxdh Kernel
>> modules.
>> +The NPSDK library and zxdh Kernel modules are not part of DPDK and
Hi Jerin,
> -Original Message-
> From: Jerin Jacob
> Sent: Thursday, June 6, 2024 3:33 PM
> To: Kundapura, Ganapati
> Cc: dev@dpdk.org; jer...@marvell.com; Gujjar, Abhinandan S
> ; Jayatheerthan, Jay
> ; Naga Harish K, S V
>
> Subject: Re: [PATCH v2] eventdev/crypto: restore opaque fiel
For session-less crypto operations, event info is contained in
crypto op metadata for each event which is restored in event
from the crypto op metadata response info.
For session based crypto operations, crypto op contains per session
based event info in crypto op metadata. If any PMD passes any
i
With modern CPUs, it is possible to have higher
CPU count thus we can have higher RTE_MAX_LCORES.
In testpmd application, the current config forwarding
cores option "--nb-cores" is hard limited to 255.
The patch fixes this constraint and also adjusts the lcore
data structure to 32-bit to align wit
>
>
> > On Jun 4, 2024, at 9:28 AM, Boyer, Andrew wrote:
> >
> >
> >> On Apr 30, 2024, at 4:21 PM, Boyer, Andrew
> wrote:
> >>
> >> This patchset introduces a new crypto PMD for AMD Pensando hardware
> >> accelerators. It allows applications running directly on the AMD Pensando
> >> DSC to offl
Hi,
From: Suanming Mou
Sent: Wednesday, June 5, 2024 12:37 PM
To: Dariusz Sosnowski; Slava Ovsiienko; Ori Kam; Matan Azrad
Cc: dev@dpdk.org; Raslan Darawsheh
Subject: [PATCH v2] net/mlx5: support match with E-Switch manager
Currently, in switch mode, mlx5 PMD only supports match with
dedicated v
Hi,
From: Suanming Mou
Sent: Wednesday, June 5, 2024 12:31 PM
To: Dariusz Sosnowski; Slava Ovsiienko; Ori Kam; Matan Azrad
Cc: dev@dpdk.org; Raslan Darawsheh
Subject: [PATCH v2 1/3] net/mlx5: add match with Tx queue item
With the item RTE_FLOW_ITEM_TYPE_TX_QUEUE, user will be able
to set the Tx
You may skip the ones which is common between kernel and DPDK and fix the ones
which are specific to DPDK.
From: Boyer, Andrew
Sent: Wednesday, June 5, 2024 6:57 AM
To: Akhil Goyal
Cc: Boyer, Andrew ; dev@dpdk.org
Subject: Re: [EXTERNAL] [PATCH v2 1/9] crypto/ionic: introduce AMD Pensando
ioni
Check whether CPFL PMD runs on Host or ACC
---
v2 Changes:
-Changed implementation based on review comment.
v3 Changes:
-Fixed indentation.
v4 Changes:
-Fix ipu_imc and ipu_acc to ipu-imc and ipu-acc.
v5 Changes:
-Updated the documentation with the changes implemented
in this patch.
---
Signed-o
> Subject: [EXTERNAL] [PATCH v3 1/5] crypto/openssl: fix GCM and CCM thread
> unsafe ctxs
>
> Commit 67ab783b5d70 ("crypto/openssl: use local copy for session
> contexts") introduced a fix for concurrency bugs which could occur when
> using one OpenSSL PMD session across multiple cores simultaneou
Hello Devs,
I hope this email finds you well.
I am reaching out to seek assistance regarding an issue I am facing in
DPDK-BBdev utility, (v22.11).
I am performing some tests over a 5G Accelerator, ACC200 using DPDK-BBdev. I am
creating custom test-vector files to generate custom load over the
Hello Devs,
I hope this email finds you well.
I am reaching out to seek assistance regarding an issue I am facing in
DPDK-BBdev utility, (v22.11).
I am performing some tests over a 5G Accelerator, ACC200 using DPDK-BBdev. I am
creating custom test-vector files to generate custom load over the
From: Yevgeny Kliteynik
HWS API is very different from SWS and has limitations:
- Queue based async insertion/deletion
- Doesn’t handle complex rules
- Requires the user to specify fixed matcher size – no rehash
- Requires the user to specify action combination on matcher creation
- Matching
From: Yevgeny Kliteynik
Add support for MLX5DR_ACTION_TYP_LAST action type.
Signed-off-by: Yevgeny Kliteynik
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/hws/mlx5dr.h| 12
drivers/net/mlx5/hws/mlx5dr_action.c | 9 +
2 files changed, 21 insertions(+)
diff --g
Initial coding to support non-template in mlxdr layer.
Yevgeny Kliteynik (3):
net/mlx5/hws: split the root rule creation and destruction
net/mlx5/hws: add support for action type LAST
net/mlx5/hws: add support for backward-compatible API
drivers/net/mlx5/hws/meson.build | 1 +
driv
From: Yevgeny Kliteynik
Split the root rule creation/destruction into two stages:
- do the job (create/destroy rule)
- generate completion
Completion generation is required for the usual HWS API.
The create/destroy functions that don't generate completion
are exposed in header file and will be
For ASO max allocations in non-template mode,
Read FW capabilities instead of using consts.
Signed-off-by: Maayan Kashani
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx5.h| 17 -
drivers/net/mlx5/mlx5_flow_hw.c| 13 +
drivers/net/mlx5/mlx5_flow_
Change flow_drv_list_create/destroy to mlx5_flow_list_create/destroy.
Remove resource release function inlining.
Check number of queues in template mode in hw configure function.
Use user priority to calculate matcher priority.
Signed-off-by: Maayan Kashani
Acked-by: Dariusz Sosnowski
---
driv
Add support for encap/decap/modify header
action for non template API.
Save 1 action per bulk according to action data.
Reuse action if possible.
Store actions same as for SWS today, use same key structure.
Signed-off-by: Maayan Kashani
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx5_flo
reuse HWS actions translation code in order to create
the actions for lower layer.
All actions are handled as non masked.
Therefore, will be translated in the rule construct
phase shared with the template code
and not during the action template translate.
Currently there is no option to reconfigur
Need to read maximum connection tracking
objects from HCA attributes.
Signed-off-by: Maayan Kashani
Acked-by: Dariusz Sosnowski
---
drivers/common/mlx5/mlx5_devx_cmds.c | 3 +++
drivers/common/mlx5/mlx5_devx_cmds.h | 1 +
2 files changed, 4 insertions(+)
diff --git a/drivers/common/mlx5/mlx5_d
If the counter pool was not added to list,
and an error state was reached,
on attempt to destroy the counter pool,
segmentation fault was received during list remove action.
Added a check to verify the list is not empty before trying to
remove the cpool from the list.
Invalid state, leading to se
This patch adds counter, connection tracking, meter
and age actions support to non-template mode.
For CT, counter and meter: if no previous allocation was
handled by hw configure routine,
Half of the maximum supported number of objects will be allocated.
For AGE action, if no counters were allocat
From: Bing Zhao
The validation will be handled later.
Now it only supports the SEND_TO_KERNEL on non-root tables.
Signed-off-by: Bing Zhao
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx5_flow_hw.c | 23 ---
1 file changed, 20 insertions(+), 3 deletions(-)
diff --gi
From: Bing Zhao
Support JUMP / DROP / QUEUE / MARK / FLAG now.
Signed-off-by: Bing Zhao
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx5_flow_hw.c | 153
1 file changed, 135 insertions(+), 18 deletions(-)
diff --git a/drivers/net/mlx5/mlx5_flow_hw.c b/dr
From: Bing Zhao
The RTE_FLOW_ACTION_TYPE_END action needs to be translated into the
MLX5DR_ACTION_TYP_LAST for the future usage. It is only needed in
the hardware steering backward compatible API.
Signed-off-by: Bing Zhao
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx5.h | 1 +
Implementing non-template rules creation supporting direct actions possible.
Bing Zhao (3):
net/mlx5: add dummy last action
net/mlx5: add basic actions support for non-template API
net/mlx5: add default miss action support in nt2hws mode
Maayan Kashani (8):
net/mlx5: initial design of non
Implement the frame and needed building
blocks for non template to hws API's.
Added validate, list_create and list_destroy to mlx5_flow_hw_drv_ops.
Rename old list_create/list_destroy functions to legacy_*
and added a call from verbs/dv ops to the legacy functions.
Updated rte_flow_hw as needed.
Setting the cipher padding has a noticeable performance footprint,
and it doesn't need to be done for every call to
process_openssl_cipher_{en,de}crypt(). Setting it causes OpenSSL to set
it on every future context re-init. Thus, for every buffer after the
first one, the padding is being set twice.
Currently EVP auth ctxs (e.g. EVP_MD_CTX, EVP_MAC_CTX) are allocated,
copied to (from openssl_session), and then freed for every auth
operation (ie. per packet). This is very inefficient, and avoidable.
Make each openssl_session hold an array of structures, containing
pointers to per-queue-pair ci
Currently EVP_CIPHER_CTXs are allocated, copied to (from
openssl_session), and then freed for every cipher operation (ie. per
packet). This is very inefficient, and avoidable.
Make each openssl_session hold an array of pointers to per-queue-pair
cipher context copies. These are populated on first
Currently the 3DES-CTR cipher context is initialised for every buffer,
setting the cipher implementation and key - even though for every
buffer in the session these values will be the same.
Change to initialising the cipher context once, before any buffers are
processed, instead.
Throughput perfo
v2:
* Fixed missing * in patch 4 causing compilation failures.
v3:
* Work around a lack of support for duplicating EVP_CIPHER_CTXs for
AES-GCM and AES-CCM in OpenSSL versions 3.0.0 <= v < 3.2.0.
---
The current implementation of the OpenSSL PMD has numerous performance issues.
These revolve aro
Commit 67ab783b5d70 ("crypto/openssl: use local copy for session
contexts") introduced a fix for concurrency bugs which could occur when
using one OpenSSL PMD session across multiple cores simultaneously. The
solution was to clone the EVP contexts per-buffer to avoid them being
used concurrently.
From: Gregory Etelson
Support non-template flows API in FDB mode:
dpdk-testpmd -a $PCI,dv_flow_en=2,representor=vf0-1 -- -i
testpmd> flow create 0 group 0 transfer \
pattern eth / end \
actions count / drop / end
Signed-off-by: Gregory Etelson
Acked-by: Dariusz Sos
From: Gregory Etelson
HWS ASO actions validation required PMD to allocate resources during
the port configuration, before the action validation was called.
That approach does not work in the HWS non-template setup, because
non-template setup does not have explicit port configuration
procedure an
From: Gregory Etelson
Non-template PMD implementation allocates ASO flow actions resources
on-demand.
Current PMD implementation iterated over actions array in search for
actions that required ASO resources allocations.
The patch replaced the iteration with actions flags bitmap queries.
Signed
From: Gregory Etelson
Add support for the RSS, AGE, COUNT and CONNTRACK indirect flow
actions for the non-template flow rules.
Signed-off-by: Gregory Etelson
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx5_flow_hw.c | 111 +---
1 file changed, 89 insertions(+
From: Gregory Etelson
The MLX5 PMD expands flow rule with the RSS action in the
non-template environment.
The patch adds RSS flow rule expansion for legacy flow rules in
the template setup.
Signed-off-by: Gregory Etelson
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/hws/mlx5dr_definer.c |
Adding FDB support, RSS expansion and indirect actions.
Gregory Etelson (6):
net/mlx5: update NTA rule pattern and actions flags
net/mlx5: support RSS expansion in non-template HWS setup
net/mlx5: support indirect actions in non-template setup
net/mlx5: update ASO resources in non-template
From: Gregory Etelson
Move pattern flags bitmap to flow_hw_list_create.
Create actions flags bitmap in flow_hw_list_create.
PMD uses pattern and actions bitmaps for direct queries instead of
iterating arrays.
Signed-off-by: Gregory Etelson
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx
review and cleanup unneeded TODO comments.
Signed-off-by: Maayan Kashani
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx5_flow_hw.c | 32 +---
1 file changed, 9 insertions(+), 23 deletions(-)
diff --git a/drivers/net/mlx5/mlx5_flow_hw.c b/drivers/net/mlx5/mlx5_
In current implementation, in non template mode,
encap action is not set as non shared(according to given masks).
By masking the relevant fields, encap is now used as shared.
decap remained unshared and cannot be shared according to
current implementation.
Optimize encap action mask initialization
In current implementation, in non template mode,
modify header action is not set as always shared.
Align to HWS implementation, setting modify header action as always shared.
Optimize mask initialization.
Signed-off-by: Maayan Kashani
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx5_flow.h
Reorganize rte_flow_hw and rte_flow_nt2hws
structures for better performance, and removed packed.
Signed-off-by: Maayan Kashani
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/mlx5_flow.h | 26 --
1 file changed, 12 insertions(+), 14 deletions(-)
diff --git a/drivers/n
Code changes applied after review.
Maayan Kashani (4):
net/mlx5: reorganize main structures
net/mlx5: set modify header as shared action
net/mlx5: set encap as shared action
net/mlx5: clean up TODO comments
drivers/net/mlx5/mlx5_flow.h| 29 ++--
drivers/net/mlx5/mlx5_flow_dv.c | 12
On Thu, Jun 6, 2024 at 1:57 PM Jerin Jacob wrote:
>
> On Tue, Jun 4, 2024 at 9:49 PM Ganapati Kundapura
> wrote:
> >
> > For session-less crypto operations, event info is contained in
> > crypto op metadata for each event which is restored in event
> > from the crypto op metadata response info.
>
On Wed, Jun 5, 2024 at 7:43 PM Mattias Rönnblom
wrote:
>
> Add the RTE_EVENT_DEV_CAP_IMPLICIT_RELEASE_DISABLE capability to the
> DSW event device.
>
> This feature may be used by an EAL thread to pull more work from the
> work scheduler, without giving up the option to forward events
> originatin
[AMD Official Use Only - AMD Internal Distribution Only]
Hi Ferruh,
> -Original Message-
> From: Yigit, Ferruh
> Sent: Friday, April 19, 2024 5:00 PM
> To: Tummala, Sivaprasad ;
> david.march...@redhat.com; aman.deep.si...@intel.com;
> yuying.zh...@intel.com
> Cc: dev@dpdk.org; sta...@dp
From: Yevgeny Kliteynik
If resized rule is deleted, the delete process also includes clearing
the rule's resize_info. The problem is, this resize_info contains info
on the action STEs that are freed onle when the completion of rule
deletion received.
This means that resize_info should be freed on
From: Yevgeny Kliteynik
Reorganize struct mlx5dr_rule_resize_info fields to
reduce its size in memory.
Signed-off-by: Yevgeny Kliteynik
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/hws/mlx5dr_rule.h | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/net/mlx5/h
From: Yevgeny Kliteynik
When destroying resizable matcher, we iterate through the list
of previous src matchers and free all their action_ste resources.
Now that we're supporting growing action template sizes, we also
need to support a case when each new dst matcher that we resize
into might have
From: Yevgeny Kliteynik
If during rehash there is a completion with error on moving
rule to a new matcher, abort the rehash and return error.
Signed-off-by: Yevgeny Kliteynik
Acked-by: Dariusz Sosnowski
---
drivers/net/mlx5/hws/mlx5dr_bwc.c | 25 ++---
1 file changed, 18 i
1 - 100 of 118 matches
Mail list logo
