On Thu, Jul 18, 2024 at 11:43:26AM -0700, Stephen Hemminger wrote:
> Many places in the documentation are using -n 4 to set
> the number of memory channels. This should not be recommended
> since it is not always right and the default should be used
> instead.
>
> Signed-off-by: Stephen Hemminger
On Thu, Jul 18, 2024 at 12:07:13PM -0700, Stephen Hemminger wrote:
> The cksum tests would not get run since not part of one
> of the test suites. Meson complains with:
>
> WARNING: Test "cksum_perf_autotest" is not defined in any test suite
>
> Signed-off-by: Stephen Hemminger
Acked-by: Bruce
On 7/17/2024 2:44 PM, Patrick Robb wrote:
> On Mon, Jul 15, 2024 at 10:38 PM zhoumin wrote:
>>
>> Hi Patrick,
>>
>> Thanks for giving the link of commit `a6c3ec342ee1`.
>>
>> However I cannot checkout this commit in the next-net repository because
>> the commit ID is not exist in the history of ne
On 7/15/2024 1:13 PM, Michael Baum wrote:
> The "rte_flow_conv()" function, enables, among other things, to copy
> item list.
>
> For GENEVE option item, the function copies it without considering deep
> copy. It copies the "data" pointer without copying the pointed values.
>
> This patch adds de
On 7/17/2024 4:24 AM, Chaoyong He wrote:
> This patch series fix two coverity issues:
> - 439964
> - 439967
>
> Chaoyong He (1):
> net/nfp: fix copy and paste error
>
> Long Wu (1):
> net/nfp: fix AVX2 vector Rx function memory overrun
>
Series applied to dpdk-next-net/main, thanks.
On 7/11/2024 1:35 PM, Bruce Richardson wrote:
> When allocating memory for an ethdev, the rte_malloc_socket call used
> only allocates memory on the NUMA node/socket local to the device. This
> means that even if the user wanted to, they could never use a remote NIC
> without also having memory on
On 7/8/2024 12:45 PM, Ferruh Yigit wrote:
> On 7/8/2024 4:39 AM, Mihai Brodschi wrote:
>>
>>
>> On 07/07/2024 21:46, Mihai Brodschi wrote:
>>>
>>>
>>> On 07/07/2024 18:18, Mihai Brodschi wrote:
On 07/07/2024 17:05, Ferruh Yigit wrote:
>
> My expectation is numbers should be l
From: Dengdui Huang
In the HNS3 driver, pkt_len in mbuf do not contain the CRC length.
This patch fix it.
Fixes: 8973d7c4ca12 ("net/hns3: support keeping CRC")
Cc: sta...@dpdk.org
Signed-off-by: Dengdui Huang
Acked-by: Huisong Li
Acked-by: Jie Hai
---
drivers/net/hns3/hns3_rxtx.c |
From: Dengdui Huang
The data exceeds the pkt_len in mbuf is inavailable for user.
When KEEP CRC offload is enabled, CRC field length should be
included in the pkt_len in mbuf. However, almost of drivers
supported KEEP CRC feature didn't add the CRC data length to
pkt_len. So it is very necessary
From: Dengdui Huang
When KEEP_CRC offload is enabled, the CRC data is still stripped
in following cases:
1. For HIP08 network engine, the packet type is TCP and the length
is less than or equal to 60B.
2. For HIP09 network engine, the packet type is IP and the length
is less than or equal t
For hns3 NIC, when KEEP_CRC offload is enabled, the CRC data is
still be stripped in rare scenarios. Some users of hns3 are
already using this feature. So driver has to recaculate packet CRC.
In addition, in the mbuf, the data that exceeds the length specified
by pkt_len is invalid. Therefore, if t
> It would be good to make sure that DPDK BPF conforms to IETF draft.
> https://datatracker.ietf.org/doc/draft-ietf-bpf-isa/
>
> Based on LWN article on presentation at Linux Storage, Filesystem,
> Memory Mangerment, and BPF Summit.
>
> https://lwn.net/SubscriberLink/975830/3b32df6be23d3abf/
> From: Robin Jarry [mailto:rja...@redhat.com]
>
> Vladimir Medvedkin, Jul 18, 2024 at 23:25:
> > I think alignment should be 1 since in FIB6 users usually don't copy
> IPv6
> > address and just provide a pointer to the memory inside the packet.
How can they do that? The bulk lookup function take
On Thu, Jul 18, 2024 at 12:07:14PM -0700, Stephen Hemminger wrote:
> The red tests were not run because not part of any suite.
> Meson warning is:
> WARNING: Test "red_autotest" is not defined in any test suite
>
> Signed-off-by: Stephen Hemminger
Acked-by: Bruce Richardson
> ---
> app/test/
On Thu, Jul 18, 2024 at 12:07:15PM -0700, Stephen Hemminger wrote:
> The tests were not part of any suite so not run normally.
> Meson warning is:
> WARNING: Test "timer_secondary_autotest" is not defined in any test suite
>
> Signed-off-by: Stephen Hemminger
> ---
Acked-by: Bruce Richardson
Series-reviewed-by: Chengwen Feng
On 2024/7/19 17:04, Jie Hai wrote:
> For hns3 NIC, when KEEP_CRC offload is enabled, the CRC data is
> still be stripped in rare scenarios. Some users of hns3 are
> already using this feature. So driver has to recaculate packet CRC.
> In addition, in the mbuf, th
On Fri, Jul 19, 2024 at 09:59:50AM +0100, Ferruh Yigit wrote:
> On 7/11/2024 1:35 PM, Bruce Richardson wrote:
> > When allocating memory for an ethdev, the rte_malloc_socket call used
> > only allocates memory on the NUMA node/socket local to the device. This
> > means that even if the user wanted
From: Jun Yang
struct {
uint64_t coreid : 4; /**--rbp.sportid / rbp.dportid*/
uint64_t pfid : 8; /**--rbp.spfid / rbp.dpfid*/
uint64_t vfen : 1; /**--rbp.svfa / rbp.dvfa*/
uint64_t vfid : 16; /**--rbp.svfid / rbp.dvfid*/
.
} pcie;
Signed-off-by: Jun Ya
From: Jun Yang
Initialize and Configure queues of dma device according to hw queues
supported from mc bus.
Because multiple queues per device are supported, virt queues
implementation are dropped.
Signed-off-by: Jun Yang
---
drivers/dma/dpaa2/dpaa2_qdma.c | 312 +++-
From: Jun Yang
1) Support DMA single copy and SG copy.
2) Silent mode support.
Add index combined with length field.
For Silent mode, this index is used to notify DMA driver
which inner descriptor should be used.
For none silent mode, this index is used to notify user
which descriptor is complet
From: Jun Yang
Support multiple processes for dpaa2 dma.
1) Move queue configuration procedure from init function to
device configuration function which is called by user.
2) Instances of dpaa2_dpdmai_dev and qdma_device are allocated
from primary process and shared between multiple processes.
From: Jun Yang
Make sure the SG entry number doesn't overflow.
Signed-off-by: Jun Yang
---
drivers/dma/dpaa2/dpaa2_qdma.c | 11 ++-
1 file changed, 10 insertions(+), 1 deletion(-)
diff --git a/drivers/dma/dpaa2/dpaa2_qdma.c b/drivers/dma/dpaa2/dpaa2_qdma.c
index 51affed40c..e521df8817
From: Jun Yang
Include dpaa2_hw_pvt.h and dpaa2_hw_dpio.h files
Signed-off-by: Jun Yang
---
drivers/dma/dpaa2/dpaa2_qdma.h | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/dma/dpaa2/dpaa2_qdma.h b/drivers/dma/dpaa2/dpaa2_qdma.h
index 743a43fa14..eb02bff08f 100644
--- a/drivers/dm
From: Jun Yang
For copy_sg: pass job index lists.
For copy: pass job index.
Signed-off-by: Jun Yang
---
drivers/dma/dpaa2/dpaa2_qdma.c | 92 ++
drivers/dma/dpaa2/dpaa2_qdma.h | 7 ++
drivers/dma/dpaa2/rte_pmd_dpaa2_qdma.h | 15 -
3 files changed, 68
From: Jun Yang
Multiple DQ storages are used among multiple cores, the single dq
storage of first union is leak if multiple storages are allocated.
It does not make sense to keep the single dq storage of union,
remove it and reuse the first dq storage of multiple storages
for this case.
Signed-o
From: Jun Yang
Short FD can be used for single transfer scenario which shows higher
performance than FLE.
1) Save index context in FD att field for short and FLE(NonSG).
2) Identify FD type by att of FD.
3) Force 48 bits address for source address and fle according to spec.
Signed-off-by: Jun Ya
From: Jun Yang
For non-SG format, the index is saved in FD with
DPAA2_QDMA_FD_ATT_TYPE_OFFSET(13) bits width.
The max descriptor number of ring is power of 2, so the
eventual max number is:
((1 << DPAA2_QDMA_FD_ATT_TYPE_OFFSET) / 2)
Signed-off-by: Jun Yang
---
drivers/dma/dpaa2/dpaa2_qdma.h |
From: Jun Yang
The return value of DMA copy/sg copy should be index of
descriptor copied in success.
Signed-off-by: Jun Yang
---
drivers/dma/dpaa2/dpaa2_qdma.c | 14 +++---
1 file changed, 11 insertions(+), 3 deletions(-)
diff --git a/drivers/dma/dpaa2/dpaa2_qdma.c b/drivers/dma/dpaa2
From: Jun Yang
Include rte_pmd_dpaax_qdma.h instead of rte_pmd_dpaa2_qdma.h
and change code accordingly.
Signed-off-by: Jun Yang
---
doc/api/doxy-api-index.md | 2 +-
doc/api/doxy-api.conf.in | 2 +-
drivers/common/dpaax/meson.build | 3 +-
drivers/
This patch add support to use multiple dma channel in the driver.
Signed-off-by: Gagandeep Singh
---
drivers/dma/dpaa/dpaa_qdma.c | 8
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/drivers/dma/dpaa/dpaa_qdma.c b/drivers/dma/dpaa/dpaa_qdma.c
index 10e65ef1d7..24ad7ad019 1
The check shall be end instead of equal.
Fixes: 7da29a644c51 ("dma/dpaa: support DMA operations")
Cc: sta...@dpdk.org
Signed-off-by: Gagandeep Singh
---
drivers/dma/dpaa/dpaa_qdma.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/dma/dpaa/dpaa_qdma.c b/drivers/dma/dp
From: Vanshika Shukla
This patch improves the dpaa qdma driver and
adds dpaa_qdma_burst_capacity API which returns the
remaining space in the descriptor ring.
Signed-off-by: Vanshika Shukla
---
drivers/dma/dpaa/dpaa_qdma.c | 34 +-
drivers/dma/dpaa/dpaa_qdma.h |
ERR050757 on LS104x indicates:
For outbound PCIe read transactions, a completion buffer is used
to store the PCIe completions till the data is passed back to the
initiator. At most 16 outstanding transactions are allowed and
maximum read request is 256 bytes. The completion buffer size
inside the
Non-prefetchable read setting in the source descriptor may be
required for targets other than local memory. Prefetchable read
setting will offer better performance for misaligned transfers
in the form of fewer transactions and should be set if possible.
This patch also fixes QDMA stall issue due to
remove unwanted descriptor list maintenance
and channels overhead.
Signed-off-by: Gagandeep Singh
---
drivers/dma/dpaa/dpaa_qdma.c | 594 +--
drivers/dma/dpaa/dpaa_qdma.h | 43 +--
2 files changed, 221 insertions(+), 416 deletions(-)
diff --git a/drivers/dma/dpa
Remove unnessary status read before every send.
Signed-off-by: Gagandeep Singh
---
drivers/dma/dpaa/dpaa_qdma.c | 186 ++-
drivers/dma/dpaa/dpaa_qdma.h | 7 ++
2 files changed, 101 insertions(+), 92 deletions(-)
diff --git a/drivers/dma/dpaa/dpaa_qdma.c b/drive
From: Jun Yang
Return complete index instead of total complete counter
in complete callback.
Signed-off-by: Jun Yang
---
drivers/dma/dpaa/dpaa_qdma.c | 534 ++-
drivers/dma/dpaa/dpaa_qdma.h | 109 +++
2 files changed, 330 insertions(+), 313 deletions(-)
dif
From: Jun Yang
To support multiple command queues of each block, status queue
need support to notify which command queue of block is completed.
The multiple command queues are balanced to blocks in setup.
If multiple command queues are enabled in one block,
de-queue status is performed instead o
From: Jun Yang
Perform SG operation by copy_sg callback of DMA lib or
burst request from application.
Perform Simple operation if burst number is 1.
Signed-off-by: Jun Yang
---
drivers/dma/dpaa/dpaa_qdma.c | 856 ++-
drivers/dma/dpaa/dpaa_qdma.h | 184 +---
From: Jun Yang
Perform block dequeue to identify which queue of this block is completed.
Signed-off-by: Jun Yang
---
drivers/dma/dpaa/dpaa_qdma.c | 252 ---
1 file changed, 116 insertions(+), 136 deletions(-)
diff --git a/drivers/dma/dpaa/dpaa_qdma.c b/drivers/
From: Jun Yang
The congestion issue occurs frequently on low speed device(PCIe).
We should drain the command queue to make dma work when congestion occurs.
Signed-off-by: Jun Yang
---
drivers/dma/dpaa/dpaa_qdma.c | 157 +++
1 file changed, 85 insertions(+), 72 d
From: Jun Yang
Burst operation used for SG copy as default until SG issue is fixed.
Signed-off-by: Jun Yang
---
drivers/dma/dpaa/dpaa_qdma.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/drivers/dma/dpaa/dpaa_qdma.c b/drivers/dma/dpaa/dpaa_qdma.c
index 93199cf9a3..1c833
From: Jun Yang
Fix issue of ERRATA 050757/050265 workaround which is not effective
in burst mode.
SDF/DDF is referred by first entry of compound frame table, move the DF to
compound frame table description which is suitable to adapt single copy
and SG/burst copy.
Fix SG issue which was caused b
From: Jun Yang
Don't save eq context in silent mode, check HW status only to
identify if queue is full.
Signed-off-by: Jun Yang
---
drivers/dma/dpaa/dpaa_qdma.c | 14 --
1 file changed, 12 insertions(+), 2 deletions(-)
diff --git a/drivers/dma/dpaa/dpaa_qdma.c b/drivers/dma/dpaa/d
From: Jun Yang
Split burst copies to multiple SG copies if burst number exceeds
max number of SG entries.
Signed-off-by: Jun Yang
---
drivers/dma/dpaa/dpaa_qdma.c | 180 +++
drivers/dma/dpaa/dpaa_qdma.h | 2 +-
2 files changed, 98 insertions(+), 84 deletions(-
From: Jun Yang
SG transaction is not helpful for performance of large transaction size.
Start single transaction for size > max SG entry size in SG copy.
Default MAX SG entry size is 2000 bytes which is experiment data of
mem to mem, user can change it according to experiment:
export DPAA_QDMA_S
From: Hemant Agrawal
Add BMI statistics and fixing the existing extented
statistics
Signed-off-by: Hemant Agrawal
Signed-off-by: Gagandeep Singh
---
drivers/bus/dpaa/base/fman/fman_hw.c | 65 +++-
drivers/bus/dpaa/include/fman.h | 4 +-
drivers/bus/dpaa/include/f
Morten Brørup, Jul 19, 2024 at 11:12:
> Vladimir Medvedkin, Jul 18, 2024 at 23:25:
> > I think alignment should be 1 since in FIB6 users usually don't
> > copy IPv6 address and just provide a pointer to the memory inside
> > the packet.
How can they do that? The bulk lookup function takes an
On Fri, Jul 19, 2024 at 12:02:38PM +0200, Robin Jarry wrote:
> Morten Brørup, Jul 19, 2024 at 11:12:
> > > Vladimir Medvedkin, Jul 18, 2024 at 23:25:
> > > > I think alignment should be 1 since in FIB6 users usually don't > >
> > copy IPv6 address and just provide a pointer to the memory inside > >
On Thu, Jul 18, 2024 at 12:07:12PM -0700, Stephen Hemminger wrote:
> This test should be using the TEST_ASSERT macros, and can be
> run as part of the fast test suite now.
>
> Signed-off-by: Stephen Hemminger
> ---
> app/test/test_alarm.c | 53 ---
> 1 fil
On Fri, Jul 19, 2024 at 09:59:50AM +0100, Ferruh Yigit wrote:
> On 7/11/2024 1:35 PM, Bruce Richardson wrote:
> > When allocating memory for an ethdev, the rte_malloc_socket call used
> > only allocates memory on the NUMA node/socket local to the device. This
> > means that even if the user wanted
Hi Morten,
On 19/07/2024 10:12, Morten Brørup wrote:
From: Robin Jarry [mailto:rja...@redhat.com]
Vladimir Medvedkin, Jul 18, 2024 at 23:25:
I think alignment should be 1 since in FIB6 users usually don't copy
IPv6
address and just provide a pointer to the memory inside the packet.
How can
> From: Robin Jarry [mailto:rja...@redhat.com]
>
> Morten Brørup, Jul 19, 2024 at 11:12:
> > > Vladimir Medvedkin, Jul 18, 2024 at 23:25:
> > > > I think alignment should be 1 since in FIB6 users usually don't
> > > > copy IPv6 address and just provide a pointer to the memory inside
> > > > the pa
Morten Brørup, Jul 19, 2024 at 12:46:
When passing an IPv4 address as a value, alignment does matter; it
must be 4 byte aligned.
I was expecting the compiler to do what is necessary to copy the data to
an aligned register before jumping to the function.
On a CPU with 128 bit registers, I wou
On 7/19/2024 10:57 AM, Bruce Richardson wrote:
> On Fri, Jul 19, 2024 at 09:59:50AM +0100, Ferruh Yigit wrote:
>> On 7/11/2024 1:35 PM, Bruce Richardson wrote:
>>> When allocating memory for an ethdev, the rte_malloc_socket call used
>>> only allocates memory on the NUMA node/socket local to the de
> Add the option to create one session for the PMD, and share it across
> all of the queue pairs. This may help to discover/debug concurrency
> issues (both correctness and performance) that can occur when using this
> configuration.
>
> Signed-off-by: Jack Bond-Preston
> Acked-by: Brian Dooley
> > This patch fixes incorrect placement of OOP offset.
> > Data preceding crypto operation is not copied to the output buffer, which is
> > conformant to the API.
> >
> > Fixes: a815a04cea05 ("crypto/qat: support symmetric build op request")
> > Cc: sta...@dpdk.org
> >
> > Signed-off-by: Arkadiusz
> Subject: RE: [PATCH v1] test/crypto: fix comparison function for modex values
>
> > Subject: [PATCH v1] test/crypto: fix comparison function for modex values
> >
> > Fix comparison function used by modex test to check from first non-zero
> > value
> > itself.
> >
> > Coverity issue: 430125
> >
Hi Kai,
Please review.
> From: Julien Hascoet
>
> The crypto scheduler session size computation was taking
> into account only the worker session sizes and not its own.
>
> Fixes: e2af4e403c1 ("crypto/scheduler: support DOCSIS security protocol")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Julien
On Fri, Jul 19, 2024 at 12:10:24PM +0100, Ferruh Yigit wrote:
> On 7/19/2024 10:57 AM, Bruce Richardson wrote:
> > On Fri, Jul 19, 2024 at 09:59:50AM +0100, Ferruh Yigit wrote:
> >> On 7/11/2024 1:35 PM, Bruce Richardson wrote:
> >>> When allocating memory for an ethdev, the rte_malloc_socket call
On Wed, Jul 17, 2024 at 4:23 PM Dean Marx wrote:
>
> added set promisc, set verbose, and port stop
> commands to testpmd shell.
>
> Signed-off-by: Dean Marx
Reviewed-by: Jeremy Spewock
On Wed, Jul 17, 2024 at 4:23 PM Dean Marx wrote:
>
> Configuration schema for the queue_start_stop suite.
>
> Signed-off-by: Dean Marx
Reviewed-by: Jeremy Spewock
On Wed, Jul 17, 2024 at 4:23 PM Dean Marx wrote:
>
> This suite tests the ability of the Poll Mode Driver to enable
> and disable Rx/Tx queues on a port.
>
> Signed-off-by: Dean Marx
Reviewed-by: Jeremy Spewock
On 18/07/2024 19:43, Stephen Hemminger wrote:
Many places in the documentation are using -n 4 to set
the number of memory channels. This should not be recommended
since it is not always right and the default should be used
instead.
Signed-off-by: Stephen Hemminger
Acked-by: Jack Bond-Preston
From: Bruce Richardson
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 56. When initializing DCF, attempt to get
link status data using both formats of the command, by overriding
the mac type to E830 on failure
Hi Suanming,
This change looks ok to me.
> -Original Message-
> From: Akhil Goyal
> Sent: Thursday, July 18, 2024 8:57 PM
> To: Suanming Mou ; Dooley, Brian
> ; Mcnamara, John ; Ji,
> Kai ; Kusztal, ArkadiuszX
> Cc: dev@dpdk.org
> Subject: RE: [PATCH v2] app/test-crypto-perf: add throug
On 7/19/2024 2:22 PM, Bruce Richardson wrote:
> On Fri, Jul 19, 2024 at 12:10:24PM +0100, Ferruh Yigit wrote:
>> On 7/19/2024 10:57 AM, Bruce Richardson wrote:
>>> On Fri, Jul 19, 2024 at 09:59:50AM +0100, Ferruh Yigit wrote:
On 7/11/2024 1:35 PM, Bruce Richardson wrote:
> When allocating
When running DTS with a git revision specified the following error occurs:
$ ./main.py --revision HEAD
[...]
FileNotFoundError: [Errno 2] No such file or directory:
'dts/'
Fix by casting DPDKGitTarball to Path, in order to stringify it correctly.
Bugzilla ID: 1496
Fixes: a23f22457dbd ("dts:
It looks like only formatting changes were picked up on this version
of this patch, so my previous comments still apply.
On Wed, Jul 17, 2024 at 4:31 PM Dean Marx wrote:
>
> Test suite for verifying VLAN filtering, stripping, and insertion
> functionality on Poll Mode Driver.
>
> Signed-off-by: Dean Marx
Reviewed-by: Jeremy Spewock
On Wed, Jul 17, 2024 at 4:31 PM Dean Marx wrote:
>
> Configuration to run vlan test suite
>
> Signed-off-by: Dean Marx
Reviewed-by: Jeremy Spewock
On Thu, Jul 11, 2024 at 3:31 PM Jeremy Spewock wrote:
>
> I think this change makes sense, and I mentioned this on Dean's patch
> that has the same change in it but I think we should have more
> discussion about which route to take with this addressing problem. It
> is definitely something that we
On Thu, Jul 18, 2024 at 3:12 PM Nicholas Pratte wrote:
>
> Several new methods have been added to TestPMDShell in order to produce
> the mac filter's individual test cases:
> - set_mac_addr
> - set_multicast_mac_addr
> - rx_vlan_add
> - rx_vlan_rm
> - vlan_filter_set_on
> - vlan_filter_set_o
I just noticed one mistake in a doc-comment. Otherwise however, good work.
Reviewed-by: Jeremy Spewock
On Thu, Jul 18, 2024 at 3:40 PM Nicholas Pratte wrote:
>
> The mac address filter test suite, whose test cases are based on old
> DTS's test cases, has been refactored to interface with the ne
> > > > > Subject: [PATCH v2] app/test-crypto-perf: add throughput OOP
> > > > > decryption
> > > > >
> > > > > During throughput running, re-filling the test data will impact
> > > > > the performance test result. So for now, to run decrypt throughput
> > > > > testing is not supported
> > > since
> From: Robin Jarry [mailto:rja...@redhat.com]
>
> Morten Brørup, Jul 19, 2024 at 12:46:
> > When passing an IPv4 address as a value, alignment does matter; it
> > must be 4 byte aligned.
>
> I was expecting the compiler to do what is necessary to copy the data to
> an aligned register before jum
On Fri, Jul 19, 2024 at 04:31:11PM +0100, Ferruh Yigit wrote:
> On 7/19/2024 2:22 PM, Bruce Richardson wrote:
> > On Fri, Jul 19, 2024 at 12:10:24PM +0100, Ferruh Yigit wrote:
> >> One option can be adding a warning log to the fallback case, saying that
> >> memory allocated from non-local socket a
On Thu, Jun 27, 2024 at 4:52 PM Patrick Robb wrote:
> #
> Minutes
>
> =
> General Announcements
> * DPDK Summit in Montreal will be September 24-25:
> https://ww
On Fri, 19 Jul 2024 17:47:47 +0200
Morten Brørup wrote:
> > From: Robin Jarry [mailto:rja...@redhat.com]
> >
> > Morten Brørup, Jul 19, 2024 at 12:46:
> > > When passing an IPv4 address as a value, alignment does matter; it
> > > must be 4 byte aligned.
> >
> > I was expecting the compiler
On Fri, 19 Jul 2024 11:17:43 +0100
Bruce Richardson wrote:
> On Thu, Jul 18, 2024 at 12:07:12PM -0700, Stephen Hemminger wrote:
> > This test should be using the TEST_ASSERT macros, and can be
> > run as part of the fast test suite now.
> >
> > Signed-off-by: Stephen Hemminger
> > ---
> > app/
On Fri, 19 Jul 2024 09:12:58 +
Konstantin Ananyev wrote:
> bpf_convert_filter() uses standard approach with XOR-ing itself:
> xor r0, r0, r0
> to reset some register values.
> Unfortunately linux verifier seems way too strict here and
> doesn't allow access to register with undefined value.
>
https://bugs.dpdk.org/show_bug.cgi?id=1498
Bug ID: 1498
Summary: warning in hns3 driver with clang and asan
Product: DPDK
Version: 24.03
Hardware: x86
OS: Linux
Status: UNCONFIRMED
Severity: normal
> This was taken from how kernel converts cBPF and the prologue it generates.
> Surprising that verifier gags?
Indeed it looks strange...
Have to admit I didn't dig into how linux converter works...
Below is the patch I used to test it from DPDK UT - might be you'll spot
something obvious here.
On 7/17/2024 6:56 PM, Joshua Washington wrote:
> This patch series adds the abiltiy to modify the ring size when using
> the GQ queue format for the GVE driver. Before this change, the GQ
> driver supported only 1024 descriptors in a ring. With this change, ring
> sizes can be as low or has as is s
On 7/19/2024 5:52 AM, priyadarshitathagat wrote:
> The RSS support for GVE allows multiple CPU cores to
> handle the rx/tx queues as pollers. This requires initializing
> the eth_dev_ops and updating the RX/TX functions for these pollers.
>
> Signed-off-by: Tathagat Priyadarshi
> Acked-by: Rushil
On 7/17/2024 8:11 AM, Gavin Li wrote:
> VXLAN-GPE support was removed from testpmd recently. Drivers which are
> not migrated are still using VXLAN-GPE in tests.
>
> This commit is to restore the support for VXLAN-GPE in testpmd.
>
> Fixes: da118115d95c ("app/testpmd: support matching any VXLAN f
On 7/18/2024 8:35 PM, lon...@linuxonhyperv.com wrote:
> From: Stephen Hemminger
>
> The current code uses unnecessary locking to set VF MTU, resulting in
> deadlock on hot add/remove path. Fix this by using rte_eth_dev_set_mtu()
> to set VF MTU.
>
> Signed-off-by: Stephen Hemminger
>
> Fixes:
https://bugs.dpdk.org/show_bug.cgi?id=1499
Bug ID: 1499
Summary: Intel uncore power library fails because of too many
frequencies
Product: DPDK
Version: 24.07
Hardware: x86
OS: Linux
Status: UNCO
> On Jul 16, 2024, at 9:27 PM, Stephen Hemminger
> wrote:
>
> On Mon, 15 Jul 2024 22:11:41 +
> Wathsala Vithanage wrote:
>
>> An application provides cache stashing hints to the ethernet devices to
>> improve memory access latencies from the CPU and the NIC. This patch
>> introduces thre
89 matches
Mail list logo
