On Wed, 30 Apr 2025 20:40:52 +0200
Morten Brørup wrote:
> There are only two thread priorities in the enum rte_thread_priority: Normal
> and Real-time Critical.
>
> I would like to poll ethdev counters, collect garbage and perform other
> jitter non-sensitive tasks in a control thread with low
There are only two thread priorities in the enum rte_thread_priority: Normal
and Real-time Critical.
I would like to poll ethdev counters, collect garbage and perform other jitter
non-sensitive tasks in a control thread with lower priority than my ordinary
control threads, so it will be preempt
Hello Harry,
I've been experimenting with lcore workers.
Please check out the new helloworld example:
https://github.com/getelson-at-mellanox/rdpdk/blob/safe-q/examples/helloworld.rs
There are 2 options for the example configuration:
1 Start RDPDK workers on the same cores as EAL:
cargo ru
Hey Stephan thanks for the quick reply,
I checked the code and also tested and it seems that it won't work because
rte_devargs_parse() saves the user provided name in devargs.name without
changing it to canonical form.
so in the end find_device still receives the short name.
We could save it in
On Wed, 30 Apr 2025 22:00:29 +0530
Prashant Upadhyaya wrote:
> On Wed, 30 Apr 2025 at 19:58, Stephen Hemminger
> wrote:
>
> > On Wed, 30 Apr 2025 13:00:53 +0530
> > Prashant Upadhyaya wrote:
> >
> > > > With DPDK on Azure, an application should never use the VF directly.
> > > > It needs to
On Wed, 30 Apr 2025 at 19:58, Stephen Hemminger
wrote:
> On Wed, 30 Apr 2025 13:00:53 +0530
> Prashant Upadhyaya wrote:
>
> > > With DPDK on Azure, an application should never use the VF directly.
> > > It needs to use either netvsc PMD which handles both the vmbus (slow
> path)
> > > and VF (fa
On Wed, 30 Apr 2025 11:12:51 +
Shani Peretz wrote:
> Hey Stephan,
> Apologize for the delayed response and appreciate your assistance.
> I prepared a document outlining the bug and its flow, along with the two
> solutions we discussed.
> I hope this document provides a comprehensive overview
On Wed, 30 Apr 2025 13:00:53 +0530
Prashant Upadhyaya wrote:
> > With DPDK on Azure, an application should never use the VF directly.
> > It needs to use either netvsc PMD which handles both the vmbus (slow path)
> > and VF (fast path) combined. Or use the older vdev_netvsc/failsafe/tap
> > combi
On Wed, 30 Apr 2025 15:48:36 +0200
Oleksandr Kolomeiets wrote:
> From: Danylo Vodopianov
>
> Sometimes during high flow learn performance two threads
> could cause deadlock using queue and mutex.
>
> learn_ignored variable sets when flow learn and gets
> when flow removes. Hence, learn_ignore
The macros __rte_packed_begin and __rte_packed_end are used instead
of #pragma pack(1) when structure or union memory should be minimized.
Signed-off-by: Oleksandr Kolomeiets
---
drivers/net/ntnic/include/hw_mod_flm_v25.h | 14 ++
drivers/net/ntnic/nthw/core/nthw_spi_v3.c | 15 +
From: Danylo Vodopianov
Calling memset before free() has no effect and will be flagged
by security parsing tools as a potential bug. None of these data
structures have sensitive information.
Signed-off-by: Danylo Vodopianov
---
drivers/net/ntnic/nthw/core/nthw_hif.c | 5 +
drivers/
From: Danylo Vodopianov
Add verification on NULL for variables before use it in the memset.
Signed-off-by: Danylo Vodopianov
---
drivers/net/ntnic/dbsconfig/ntnic_dbsconfig.c | 3 +++
.../link_mgmt/link_100g/nt4ga_link_100g.c | 2 ++
.../net/ntnic/nthw/flow_api/flow_id_table.c| 3
From: Danylo Vodopianov
User should call rte_ring_dequeue_zc_elem_finish to complete the
dequeue operation. However this complete step if record absent.
This change rework queue read operation.
Fixes: 96c8249be53e ("net/ntnic: learn flow queue handling")
Signed-off-by: Danylo Vodopianov
---
From: Danylo Vodopianov
There is no need to cast memset to void.
Signed-off-by: Danylo Vodopianov
---
drivers/net/ntnic/nim/i2c_nim.c | 2 +-
drivers/net/ntnic/nthw/flow_filter/flow_nthw_cat.c| 4 ++--
drivers/net/ntnic/nthw/flow_filter/flow_nthw_csu.c| 4 ++--
dr
From: Danylo Vodopianov
Sometimes during high flow learn performance two threads
could cause deadlock using queue and mutex.
learn_ignored variable sets when flow learn and gets
when flow removes. Hence, learn_ignore should not be protected
as far as two thread could not access to the same flow
From: Danylo Vodopianov
Fix CID issue 457736 divide by 0. It was avoided by
adding adittional checking on the return value
for function.
Fixes: effa04693274 ("net/ntnic: add statistics")
Signed-off-by: Danylo Vodopianov
---
drivers/net/ntnic/nthw/stat/nthw_stat.c | 4
1 file changed, 4 i
From: Danylo Vodopianov
Array usage made more explicitly, avoid pointer arithmetic.
Signed-off-by: Danylo Vodopianov
---
.../ntnic/link_mgmt/link_agx_100g/nt4ga_agx_link_100g.c | 8
drivers/net/ntnic/ntnic_ethdev.c | 2 +-
2 files changed, 5 insertions(+), 5
From: Danylo Vodopianov
Some variables set twice in the code.
Fix host side FEC enablement
Signed-off-by: Danylo Vodopianov
---
drivers/net/ntnic/nim/i2c_nim.c | 2 +-
drivers/net/ntnic/nthw/core/nthw_phy_tile.c | 1 -
drivers/net/ntnic/ntnic_vfio.c | 1 -
3 files chan
From: Danylo Vodopianov
Remove double if statement branching as far as it has the same fillings.
Signed-off-by: Danylo Vodopianov
---
.../link_agx_100g/nt4ga_agx_link_100g.c | 22 ++-
.../core/nt200a0x/reset/nthw_fpga_rst9563.c | 1 -
drivers/net/ntnic/nthw/nthw_rac.c
From: Danylo Vodopianov
Avoid usage before check
Signed-off-by: Danylo Vodopianov
---
drivers/net/ntnic/adapter/nt4ga_stat/nt4ga_stat.c | 4 ++--
drivers/net/ntnic/dbsconfig/ntnic_dbsconfig.c | 4 ++--
.../net/ntnic/link_mgmt/link_100g/nt4ga_link_100g.c | 2 +-
.../link_mgmt/l
From: Danylo Vodopianov
Add separate return value verification and logs for function.
Signed-off-by: Danylo Vodopianov
---
.../nthw/core/nt200a0x/reset/nthw_fpga_rst_nt200a0x.c| 9 +
1 file changed, 9 insertions(+)
diff --git
a/drivers/net/ntnic/nthw/core/nt200a0x/reset/nthw_fpga
From: Danylo Vodopianov
Updated logging format specifiers for better consistency
and to match the expected data types
Signed-off-by: Danylo Vodopianov
---
drivers/net/ntnic/adapter/nt4ga_adapter.c | 7 +-
drivers/net/ntnic/dbsconfig/ntnic_dbsconfig.c | 2 +-
.../link_agx_100g/nt4ga_agx_l
From: Danylo Vodopianov
Added checks for `calloc` return values to handle memory
allocation failures.
Check for null in rpp_lr_nthw_init
Signed-off-by: Danylo Vodopianov
---
drivers/net/ntnic/nthw/flow_api/flow_group.c | 17 ++--
drivers/net/ntnic/nthw/flow_api/flow_km.c | 4 +++
VLAs with known upper limits are replaced with fixed-length arrays.
In case an upper limit of an array cannot be known in advance,
the memory for such an array is allocated dynamically using alloca.
In comparison to malloc family of functions, the alloca executes faster
due to working directly with
From: Danylo Vodopianov
The conditional expressions that always has unchanging state was removed
Signed-off-by: Danylo Vodopianov
---
drivers/net/ntnic/link_mgmt/link_100g/nt4ga_link_100g.c | 2 +-
drivers/net/ntnic/nthw/core/nthw_fpga.c | 4 +---
2 files changed, 2 insertions(
From: Danylo Vodopianov
Add unsigned long long type to the value to ensure that
the value 1 is explicitly treated as a 64-bit unsigned integer.
Signed-off-by: Danylo Vodopianov
---
.../nthw/flow_api/profile_inline/flow_api_profile_inline.c | 2 +-
1 file changed, 1 insertion(+), 1 deletio
From: Danylo Vodopianov
Enhance error logging in hw_db_inline_hsh_deref function to ensure proper
handling of HSH RCP preset failures.
Signed-off-by: Danylo Vodopianov
---
.../nthw/flow_api/profile_inline/flow_api_hw_db_inline.c | 5 -
1 file changed, 4 insertions(+), 1 deletion(-)
di
From: Danylo Vodopianov
simplify conditionals and remove redundant checks
Signed-off-by: Danylo Vodopianov
---
drivers/net/ntnic/dbsconfig/ntnic_dbsconfig.c | 3 +-
.../link_agx_100g/nt4ga_agx_link_100g.c | 16 +--
drivers/net/ntnic/nim/i2c_nim.c | 3 +-
.../core/
From: Danylo Vodopianov
Remove unused CRC calculation and simplify AVR probe logging.
The condition is always false.
Signed-off-by: Danylo Vodopianov
---
drivers/net/ntnic/nthw/core/nthw_fpga.c | 299 +---
1 file changed, 11 insertions(+), 288 deletions(-)
diff --git a/dri
While strlcpy prevents out-of-bounds access,
it allows bugs to remain unnoticed.
If unexcepted truncation took place,
the respective warning message is emitted.
Signed-off-by: Oleksandr Kolomeiets
---
drivers/net/ntnic/ntnic_xstats/ntnic_xstats.c | 9 +++--
1 file changed, 7 insertions(+), 2
In the context of the NTNIC driver,
the variables named if_index stand for the index of the PHY.
However, the term if_index could also refer to the interface index
assigned by the OS.
To mitigate potential confusion these variables were renamed.
Signed-off-by: Oleksandr Kolomeiets
---
drivers/ne
The structure virtq_desc should not contain any paddings,
so it has been wrapped with __rte_packed_begin and __rte_packed_end macros.
Signed-off-by: Oleksandr Kolomeiets
---
drivers/net/ntnic/include/ntnic_virt_queue.h | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drive
The function rte_eth_stats_get fills the structure stats with zeroes
before it eventually calls dpdk_stats_collect.
That makes the call to memset inside dpdk_stats_collect redundant.
Signed-off-by: Oleksandr Kolomeiets
---
drivers/net/ntnic/ntnic_ethdev.c | 2 --
1 file changed, 2 deletions(-)
Queues with indices greater than or equal to RTE_ETHDEV_QUEUE_STAT_CNTRS
are ignored when calculating per-queue statistics,
however they should be taken into account for calculating the total.
Signed-off-by: Oleksandr Kolomeiets
---
drivers/net/ntnic/ntnic_ethdev.c | 32 ++---
* Fix issues detected by static code analyzers CoverityScan and PVS studio
- PVS studio:
https://inbox.dpdk.org/dev/caoavg170qdnxdfstjcvjhzafy-fs8e98xpba3q8u--wx+k_...@mail.gmail.com/1-a.txt
* Fix requested issues mentioned here
https://inbox.dpdk.org/dev/20250222134157.73763253@hermes.local/
Hey Stephan,
Apologize for the delayed response and appreciate your assistance.
I prepared a document outlining the bug and its flow, along with the two
solutions we discussed.
I hope this document provides a comprehensive overview. Can you take a look?
Thank you!
https://docs.google.com/docume
Pattern template validation did not take into account
the configured tunnel mode of the flex item.
This caused validation failures for templates with
flex item registered as tunnel headers.
Specifically, any item after such flex item was still considered
as outer header, not an inner.
This patch f
From: Kiran Kumar K
Adding support for IB_BTH header for RSS in cnxk device.
Signed-off-by: Kiran Kumar K
---
drivers/common/cnxk/roc_mbox.h | 1 +
drivers/net/cnxk/cnxk_ethdev.c | 3 +++
drivers/net/cnxk/cnxk_ethdev.h | 2 +-
3 files changed, 5 insertions(+), 1 deletion(-)
diff --git a/drive
From: Kiran Kumar K
On supporting hardware, RoCEv2 header can be used to
perform RSS in the ingress path.
Signed-off-by: Kiran Kumar K
---
* v4 Changes:
- Fix checkpatch issues
app/test-pmd/config.c | 5 -
doc/guides/rel_notes/release_25_07.rst | 3 +++
lib/ethdev/rte_et
From: Kiran Kumar K
Adding support for IB_BTH header for RSS in cnxk device.
Signed-off-by: Kiran Kumar K
---
drivers/common/cnxk/roc_mbox.h | 1 +
drivers/net/cnxk/cnxk_ethdev.c | 3 +++
drivers/net/cnxk/cnxk_ethdev.h | 2 +-
3 files changed, 5 insertions(+), 1 deletion(-)
diff --git a/drive
From: Kiran Kumar K
On supporting hardware, RoCEv2 header can be used to
perform RSS in the ingress path.
Signed-off-by: Kiran Kumar K
---
V3 Changes:
* Adopted review changes related to formating
app/test-pmd/config.c | 5 -
doc/guides/rel_notes/release_25_07.rst | 3 +++
On Tue, 29 Apr 2025 at 21:54, Stephen Hemminger
wrote:
> On Tue, 29 Apr 2025 21:45:36 +0530
> Prashant Upadhyaya wrote:
>
> > On Mon, 28 Apr 2025 at 21:07, Stephen Hemminger <
> step...@networkplumber.org>
> > wrote:
> >
> > > On Mon, 28 Apr 2025 11:31:10 +0530
> > > Prashant Upadhyaya wrote:
>
> From: Sunil Kumar Kori
> Sent: Tuesday, 29 April 2025 20.12
>
> rte_eth_fp_ops contains ops for fast path APIs. Each API
> validates availability of callback and then invoke it.
>
> Removing these NULL checks instead using dummy callbacks.
The description should mention the motivation for thi
43 matches
Mail list logo
