From: Anoob Joseph
Submit two instructions in one LMTLINE.
Signed-off-by: Anoob Joseph
---
drivers/common/cnxk/roc_cpt.c | 17 +-
drivers/common/cnxk/roc_cpt.h | 8 +-
drivers/crypto/cnxk/cn10k_cryptodev_ops.c | 182 +-
drivers/crypto/cnxk/cn10k_c
From: Anoob Joseph
Make 'pack_iv' variable as const to avoid multiple checks.
Signed-off-by: Anoob Joseph
---
drivers/crypto/cnxk/cnxk_se.h | 27 +--
1 file changed, 13 insertions(+), 14 deletions(-)
diff --git a/drivers/crypto/cnxk/cnxk_se.h b/drivers/crypto/cnxk/cnxk
For aes-gcm (AEAD) zero length input, sg code path is taken unlike
the digest only cases as AAD is treated as a separate input component.
Fix the zero len case in SG path by avoiding the gather component
only when it is a non AEAD algorithm. Also add sg version check as
the fix only applies to spec
From: Anoob Joseph
Add dual submission to CPT in Rx inject path.
Signed-off-by: Anoob Joseph
Signed-off-by: Vidya Sagar Velumuri
---
drivers/common/cnxk/roc_cpt.h | 43 +-
drivers/crypto/cnxk/cn10k_cryptodev_ops.c | 70 +--
drivers/crypto/cnxk/cnxk_
From: Anoob Joseph
Update sess pointer while working on next set of packets.
Signed-off-by: Anoob Joseph
---
drivers/crypto/cnxk/cn10k_cryptodev_ops.c | 13 -
1 file changed, 8 insertions(+), 5 deletions(-)
diff --git a/drivers/crypto/cnxk/cn10k_cryptodev_ops.c
b/drivers/crypto/c
From: Anoob Joseph
The packet would be posted to CPT only when there is a valid result.
Skip setting of the same.
Signed-off-by: Anoob Joseph
---
drivers/crypto/cnxk/cn10k_cryptodev_ops.c | 7 ---
1 file changed, 7 deletions(-)
diff --git a/drivers/crypto/cnxk/cn10k_cryptodev_ops.c
b/dri
From: Anoob Joseph
Use NEON instructions for Rx inject instruction preparation.
Signed-off-by: Anoob Joseph
Signed-off-by: Aakash Sasidharan
---
drivers/crypto/cnxk/cn10k_cryptodev_ops.c | 62 +--
1 file changed, 46 insertions(+), 16 deletions(-)
diff --git a/drivers/cryp
From: Anoob Joseph
RVU PF FUNC of the CPT LF need not be set as the hardware would
determine that. Instead SSO PF FUNC need to be set as inline device so
that critical errors would reach inline device.
Signed-off-by: Anoob Joseph
---
drivers/crypto/cnxk/cn10k_cryptodev_ops.c | 2 +-
drivers/cr
From: Anoob Joseph
Add flow control in Rx inject path to avoid over submission to CPT.
Signed-off-by: Anoob Joseph
---
drivers/crypto/cnxk/cn10k_cryptodev_ops.c | 16 +++-
1 file changed, 15 insertions(+), 1 deletion(-)
diff --git a/drivers/crypto/cnxk/cn10k_cryptodev_ops.c
b/dri
From: Anoob Joseph
Inline PF FUNC would be required to set SSO_PF_FUNC in the instruction
for cryptodev Rx inject. Move the API to idev to allow usage of the
same.
Signed-off-by: Anoob Joseph
---
drivers/common/cnxk/roc_idev.c | 6 ++
drivers/common/cnxk/roc_idev.h | 2
From: Anoob Joseph
Update version map with details of PMD APIs added.
Signed-off-by: Anoob Joseph
---
drivers/crypto/cnxk/rte_pmd_cnxk_crypto.h | 2 ++
drivers/crypto/cnxk/version.map | 8
2 files changed, 10 insertions(+)
diff --git a/drivers/crypto/cnxk/rte_pmd_cnxk_crypt
From: Anoob Joseph
Add comments to denote unused table entries.
Signed-off-by: Anoob Joseph
---
drivers/common/cnxk/roc_ae.c| 6 +++---
drivers/common/cnxk/roc_ae_fpm_tables.c | 6 +++---
2 files changed, 6 insertions(+), 6 deletions(-)
diff --git a/drivers/common/cnxk/roc_ae.c b/
v2:
Fix compilation errors observed with arm gcc-13.
This series adds improvements to CNXK crypto PMD and fixes aes-gcm zero
length input failure.
Aakash Sasidharan (1):
crypto/cnxk: fix aes-gcm zero len input cases
Anoob Joseph (11):
common/cnxk: add comments to denote skipped entries
cry
1. GRE_OPTION flow item validation required both item spec and mask.
HWS pattern template provides item mask only.
The patch removes spec requirement in HWS GRE_OPTION item validation.
2. In non-HWS setup, GRE_OPTION flow item validation for the checksum and
sequence item parameters require grou
From: Long Wu
The testpmd application can not modify the value of
dedicated hardware Rx/Tx queue size, and hardcoded
them as (128/512). This will cause the bonding port
start fail if some NIC requires more Rx/Tx descriptors
than the hardcoded number.
Therefore, add a command into testpmd applica
From: Long Wu
According to the check rules in the patch check script,
drivers and libraries must use the logging framework.
So standard the log message of bonding driver by using
the logging framework.
Signed-off-by: Long Wu
Reviewed-by: Peng Zhang
Reviewed-by: Chaoyong He
---
drivers/net/b
This patch series mainly add a function to bonding PMD to set
dedicated queue size, also add a command to testpmd application
to invoke this function.
At the same time, standard the log message of bonding PMD.
Long Wu (2):
net/bonding: standard the log message
net/bonding: add command to set
Based on the logic of 'bpf_parse_flags()' and the 'bpf-load' section of
'testpmd_funcs.rst' document, the 'B' flag in the help string of
bpf-load command should be '-'.
Fixes: e977e4199a8d ("app/testpmd: add commands to load/unload BPF filters")
Cc: konstantin.anan...@intel.com
Cc: sta...@dpdk.org
From: Long Wu
The app will be destroyed if add a port flow action which its
destination port id is invalid.
Add a check to avoid the problem.
Cc: chaoyong...@corigine.com
Cc: sta...@dpdk.org
Signed-off-by: Long Wu
---
drivers/net/nfp/flower/nfp_flower_flow.c | 4 ++--
1 file changed, 2 inser
CI found null pointer dereferences problem.
Coverity issue: 426271
Fixes: 153ef2e49ed7 ("net/nfp: add process private structure")
Cc: sta...@dpdk.org
Signed-off-by: Chaoyong He
Reviewed-by: Long Wu
Reviewed-by: Peng Zhang
---
drivers/net/nfp/nfp_net_common.c | 4 +++-
1 file changed, 3 insert
From: Peng Zhang
The NFP chip-wide port index is 'eth_port->index'.
In the normal card, it uses 'eth_port->eth_index', but because
'eth_port->eth_index' and 'eth_port->index' have the same value,
it doesn't have any problem.
But in the reverse sequence card, the value is different,
so it will lea
From: Peng Zhang
The logic in NFDk datapatch can make sure there only one layer VLAN,
so remove the unneeded logic here and make it clean and easy to
understand.
Fixes: 7c82b8626af8 ("net/nfp: support VLAN insert with NFDk")
Cc: sta...@dpdk.org
Signed-off-by: Peng Zhang
Reviewed-by: Chaoyong H
From: Peng Zhang
The description about the metadata of Tx direction
when using the firmware with NFDk is not correct.
Because of the exist of 8-bits metadata length, the
32-bits field type header contains only 6 4-bits
datatype specifiers rather than 8.
Fixes: c4b058784e75 ("net/nfp: modify pars
From: Zerun Fu
The original logic try to get the NSP resource every time the application
call 'rte_eth_dev_fw_version_get()' interface, there are chances the NSP
resource busy and fail to get it. And because the local string variables
not initialized, there will be out of range problem when it fa
From: Long Wu
The flow position acquire and free interfaces use different index,
and this will cause the flow free logic failed.
Fix this problem by using the same flow position index.
Fixes: 0b9079d24808 ("net/nfp: support flow API for CoreNIC firmware")
Cc: chaoyong...@corigine.com
Cc: sta...
From: Long Wu
The flow rules limitation is 2056 and driver needs
to store it in hash table. Considering hash conflicts,
driver use a new macro to enlarge hash table size.
Signed-off-by: Long Wu
Reviewed-by: Chaoyong He
Reviewed-by: Peng Zhang
---
drivers/net/nfp/nfp_net_common.h | 3 +++
dri
From: Long Wu
The firmware rules limitation is enlarged to
2056(8 Eth + 1024 IPv4 + 1024 IPv6 rules).
So enlarge the flow rules limitation and add
more precise checks to flow count.
Signed-off-by: Long Wu
Reviewed-by: Chaoyong He
Reviewed-by: Peng Zhang
---
drivers/net/nfp/nfp_net_common.h |
From: Long Wu
The data type of 'mark_id' field in 'struct nfp_net_cmsg_action'
should be 'uint32_t' rather than 'uint16_t'.
Fixes: 0b9079d24808 ("net/nfp: support flow API for CoreNIC firmware")
Cc: chaoyong...@corigine.com
Cc: sta...@dpdk.org
Signed-off-by: Long Wu
Reviewed-by: Chaoyong He
R
From: Long Wu
The port and port mask field of 'struct nfp_net_cmsg_match_v4'
are in the wrong position, and this will cause firmware not
work as expected.
Fixes: 42eabda06b0f ("net/nfp: support IPv4 flow item")
Cc: chaoyong...@corigine.com
Cc: sta...@dpdk.org
Signed-off-by: Long Wu
Reviewed-by
Remove the redundancy call of 'rte_eth_copy_pci_info()' as the API
'rte_eth_dev_pci_allocate()' already done that.
Fixes: e1124c4f8a45 ("net/nfp: add flower representor framework")
Cc: sta...@dpdk.org
Signed-off-by: Chaoyong He
Reviewed-by: Long Wu
Reviewed-by: Peng Zhang
---
drivers/net/nfp/
The original logic allow offload flow rules with empty action list, but
the matched packets will be drop by the flower firmware.
Fix this by forbidding offload this type flow rules.
Fixes: 4d946034bf9c ("net/nfp: support basic flow actions")
Cc: sta...@dpdk.org
Signed-off-by: Chaoyong He
Review
The logic of 'nfp_fw_setup()' consider both single-pf and multi-pf
firmware, but the abnormal firmware cleanup logic only consider multi-pf
firmware and try to write the heart beat value, which will cause
coredump when using the single-pf firmware.
Fixes: 8ba461d1eecc ("net/nfp: introduce keepaliv
From: Zerun Fu
In the previous logic, the link status will shown as down wrongly
whenever the call to 'nfp_eth_read_ports()' fail.
Fix this by removing the exit logic when call to 'nfp_eth_read_ports()'
fail and just giving a warning.
Fixes: 8412feed3f26 ("net/nfp: modify link update function")
From: Zerun Fu
The previous logic will call 'nfp_eth_read_ports()' every time when the
link status changed, but some of which are not necessary.
Refactor this part of logic and only call 'nfp_eth_read_ports()' when
the speed is really updated or the device is under auto-negotiation mode.
Signed
From: Zerun Fu
The speed configuration logic will execute every time the
device start, even if the speed is not changed. Refactor
this part of logic to make sure it only executes when the
speed is really need to be changed.
Signed-off-by: Zerun Fu
Reviewed-by: Chaoyong He
Reviewed-by: Long Wu
This patch series consist a branch of bugfix patches from NFP PMD,
in order to speed up the review process and avoid send them out one
by one.
Chaoyong He (4):
net/nfp: fix coredump caused by firmware abnormal cleanup
net/nfp: forbid offload flow rules with empty action list
net/nfp: remove
as sizeof(config.rx.mempool_name) is < sizeof(res->mempool), it's safer
to copy min size of them to avoide out of bound memory write.
Coverity issue: 415430
Fixes: 3850cb06ab9c ("app/graph: add ethdev commands")
Cc: sk...@marvell.com
Signed-off-by: Mahmoud Maatuq
---
app/graph/ethdev.c | 3 ++-
2024-06-23 13:30 (UTC+0100), luca.bocca...@gmail.com:
> From: Luca Boccassi
>
> We now require Meson 0.53 or later, so we can use this feature introduced
> in 0.51. This also fixes a build failure on SUSE Leap 15.6.
>
> Cc: sta...@dpdk.org
>
> Signed-off-by: Luca Boccassi
Acked-by: Dmitry Koz
From: Luca Boccassi
We now require Meson 0.53 or later, so we can use this feature introduced
in 0.51. This also fixes a build failure on SUSE Leap 15.6.
Cc: sta...@dpdk.org
Signed-off-by: Luca Boccassi
---
buildtools/meson.build | 19 +--
1 file changed, 5 insertions(+), 14 d
Hi, Igor
Thank you for the v2. The patch looks good to me, please see my further
comments below.
> > 1. The absolute max descriptor number supported by ConnectX hardware is
> 32768.
> > 2. The actual max descriptor number supported by the port (and its related
> representors)
> > reported i
Coverity issue: 426424
Fixes: e78e5408da89 ("net/mlx5: remove cache term from the list utility")
Cc: ma...@nvidia.com
Signed-off-by: Mahmoud Maatuq
---
drivers/net/mlx5/mlx5_flow_dv.c | 5 -
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/drivers/net/mlx5/mlx5_flow_dv.c b/driver
41 matches
Mail list logo