The master device and VF representors were distinguished by
presence of port name, master device did not have one. The new Linux
kernels starting from 5.0 provide the port name for master device
and the implemented representor recognizing method does not work.
The new recognizing method is based on
The mlx5 PMD probes the Verbs flow priorities supported with
ibv_create_flow() function. If rdma-core or kernel fails for
some reason, the returned error causes the drop queue is not
destroyed, and pd is locked by not freed resource.
Also the mlx5_flow_discover_priorities() returned negative value
Thursday, February 14, 2019 12:22 PM, Gaëtan Rivet:
> Subject: Re: [PATCH 5/6] net/mlx5: support PCI device DMA map and unmap
>
> On Wed, Feb 13, 2019 at 07:11:35PM +, Shahaf Shuler wrote:
> > Wednesday, February 13, 2019 1:44 PM, Gaëtan Rivet:
> > > Subject: Re: [PATCH 5/6] net/mlx5: support
The private structure stored in rte_eth_dev->data->dev_private
was named "struct priv".
In order to ease code browsing, the structure is renamed
"struct mlx[45]_priv".
Signed-off-by: Thomas Monjalon
---
drivers/net/mlx4/mlx4.c| 10 ++---
drivers/net/mlx4/mlx4.h| 20 +-
Hi,
+Shared memory packet interface (memif) PMD allows for DPDK and any other
client
+using memif (DPDK, VPP, libmemif) toc ommunicate using shared memory.
Memif is
+Linux only.s
Minor nit.
should be "to communicate"
And Also
Linux only.s => Linux only.
Reviewed-by: Rami Rosen
On 2/19/19 11:59 AM, Tiwei Bie wrote:
We should try to cleanup at least the 'need' number of descs.
Fixes: 892dc798fa9c ("net/virtio: implement Tx path for packed queues")
Cc: sta...@dpdk.org
Signed-off-by: Tiwei Bie
---
drivers/net/virtio/virtio_rxtx.c | 1 -
1 file changed, 1 deletion(
On 2/19/19 11:59 AM, Tiwei Bie wrote:
When IN_ORDER feature is negotiated, device may just write out a
single used ring entry for a batch of buffers:
"""
Some devices always use descriptors in the same order in which they
have been made available. These devices can offer the VIRTIO_F_IN_ORDER
On 2/19/19 11:59 AM, Tiwei Bie wrote:
This patch introduces a helper for clearing the virtio net header
to avoid the code duplication. Macro is used as it shows slightly
better performance.
Signed-off-by: Tiwei Bie
---
drivers/net/virtio/virtio_rxtx.c | 46 +---
On 2/19/19 11:59 AM, Tiwei Bie wrote:
When IN_ORDER feature is negotiated, device may just write out a
single used descriptor for a batch of buffers:
"""
Some devices always use descriptors in the same order in which they
have been made available. These devices can offer the VIRTIO_F_IN_ORDER
On 2/19/19 11:59 AM, Tiwei Bie wrote:
This patch introduces an optimized enqueue function in packed
ring for the case that virtio net header can be prepended to
the unchained mbuf.
Signed-off-by: Tiwei Bie
---
drivers/net/virtio/virtio_rxtx.c | 63 +++-
1 file
On Thu, Feb 21, 2019 at 12:22:29PM +0100, Maxime Coquelin wrote:
> On 2/19/19 11:59 AM, Tiwei Bie wrote:
> > This patch introduces an optimized enqueue function in packed
> > ring for the case that virtio net header can be prepended to
> > the unchained mbuf.
> >
> > Signed-off-by: Tiwei Bie
> >
On 2/21/19 1:25 PM, Tiwei Bie wrote:
On Thu, Feb 21, 2019 at 12:22:29PM +0100, Maxime Coquelin wrote:
On 2/19/19 11:59 AM, Tiwei Bie wrote:
This patch introduces an optimized enqueue function in packed
ring for the case that virtio net header can be prepended to
the unchained mbuf.
Signed-o
> On Feb 21, 2019, at 1:29 AM, Thomas Monjalon wrote:
>
> The private structure stored in rte_eth_dev->data->dev_private
> was named "struct priv".
> In order to ease code browsing, the structure is renamed
> "struct mlx[45]_priv".
>
> Signed-off-by: Thomas Monjalon
> ---
Please add "Cc: sta..
As those have been replaced by rte_bus_dma_map and rte_pci_dma_unmap
APIs.
Signed-off-by: Shahaf Shuler
---
doc/guides/prog_guide/env_abstraction_layer.rst | 2 +-
doc/guides/rel_notes/deprecation.rst| 4
lib/librte_eal/common/include/rte_vfio.h| 6 --
3 files change
Currently vfio DMA map function will fail in case the same memory
segment is mapped twice.
This is too strict, as this is not an error to map the same memory
twice.
Instead, use the kernel return value to detect such state and have the
DMA function to return as successful.
For type1 mapping the
The DPDK APIs expose 3 different modes to work with memory used for DMA:
1. Use the DPDK owned memory (backed by the DPDK provided hugepages).
This memory is allocated by the DPDK libraries, included in the DPDK
memory system (memseg lists) and automatically DMA mapped by the DPDK
layers.
2. Use
The implementation reuses the external memory registration work done by
commit[1].
Note about representors:
The current representor design will not work
with those map and unmap functions. The reason is that for representors
we have multiple IB devices share the same PCI function, so mapping will
On 2/9/2019 1:16 PM, Chas Williams wrote:
>
>
> On 1/10/19 5:22 AM, Hyong Youb Kim wrote:
>> For active-backup, tlb, and alb mode,
>> bond_ethdev_promiscuous_{enable,disable} tries to set promisc mode on
>> the primary port, even when there are no slaves. It is harmless, as
>> rte_eth_promiscuous
Enable users the option to call rte_vfio_dma_map with request to map
to the default vfio fd.
Signed-off-by: Shahaf Shuler
---
doc/guides/rel_notes/release_19_05.rst | 3 +++
lib/librte_eal/common/include/rte_vfio.h | 8 ++--
lib/librte_eal/linuxapp/eal/eal_vfio.c | 10 --
3 fil
The DPDK APIs expose 3 different modes to work with memory used for DMA:
1. Use the DPDK owned memory (backed by the DPDK provided hugepages).
This memory is allocated by the DPDK libraries, included in the DPDK
memory system (memseg lists) and automatically DMA mapped by the DPDK
layers.
2. Use
Move the memory region creation to a separate function to
prepare the ground for the reuse of it on the PCI driver map and unmap
functions.
Signed-off-by: Shahaf Shuler
---
drivers/net/mlx5/mlx5_mr.c | 86 +++--
1 file changed, 57 insertions(+), 29 deletions(-
On Thu, Feb 21, 2019 at 3:51 PM Shahaf Shuler wrote:
> As those have been replaced by rte_bus_dma_map and rte_pci_dma_unmap
> APIs.
>
> Signed-off-by: Shahaf Shuler
> ---
> doc/guides/prog_guide/env_abstraction_layer.rst | 2 +-
> doc/guides/rel_notes/deprecation.rst| 4
> lib/
Minutes 21 February 2019
Agenda:
* Release Dates
* Subtrees
* OvS
* Opens
Participants:
* Arm
* Debian
* Intel
* Mellanox
* NXP
* RedHat
Release Dates
-
* v19.05 dates, no change:
* Proposal Deadline: Friday, 1 March
* RC1: Friday, 29 March
* Release
21/02/2019 17:09, Ferruh Yigit:
> * v19.08 proposal, please comment:
> * Proposal/V1 Friday 07 June 2019
> * Integration/Merge/RC1 Friday 05 July 2019
> * Release Thurs 01 August 2019
Fixes usually take longer during July, so we may be short of t
On 2/20/2019 9:06 PM, Julien Meunier wrote:
> If the port has received less than ``pkt_per_port`` packets (for
> example, the port has missed some packets), the test is in an infinite
> loop.
>
> Instead of expecting a number of packet to receive, let the port to be
> drained by itself. If no more
> -Original Message-
> From: dev [mailto:dev-boun...@dpdk.org] On Behalf Of Anoob Joseph
> Sent: Wednesday, February 20, 2019 3:42 PM
> To: Akhil Goyal ; Doherty, Declan
> ; De Lara Guarch,
> Pablo
> Cc: Jerin Jacob Kollanukkaran ; Narayana Prasad Raju
> Athreya
> ; dev@dpdk.org; Anku
On Thu, 2019-02-21 at 16:09 +, Ferruh Yigit wrote:
> Minutes 21 February 2019
>
cut
> * Coverity is partially back, but new analyses aren't running
> * Looking for suggestions from the community for a better or more
> stable
> solutions
The clang-based alternativ
On 2/19/19 11:59 AM, Tiwei Bie wrote:
Below is a quick (unofficial) performance test (macfwd loop, 64B)
for the packed ring optimizations in this series on an Intel(R)
Xeon(R) Gold 6140 CPU @ 2.30GHz platform:
w/o this series:
packed ring normal/in-order: ~10.4 Mpps
w/ this series:
packed r
On 1/28/19 7:55 AM, sunwenjie wrote:
When rte_vhost_driver_unregister delete the connection fd,
fdset_try_del will always try and donot release the
vhostuser.mutex if the fd is busy, but the fdset_event_dispatch
will set the fd to busy and call vhost_user_msg_handler to get
vhostuser.mutex,
21/02/2019 18:43, Luca Boccassi:
> On Thu, 2019-02-21 at 16:09 +, Ferruh Yigit wrote:
> > * Looking for suggestions from the community for a better or more
> > stable solutions
>
> The clang-based alternative I mentioned that can be used from
> Travis/Github is sonarcloud.io:
>
> https://so
On 2/4/19 8:28 AM, Pallantla Poornima wrote:
sprintf function is not secure as it doesn't check the length of string.
More secure function snprintf is used.
Fixes: d7280c9fff ("vhost: support selective datapath")
Cc: sta...@dpdk.org
Signed-off-by: Pallantla Poornima
---
lib/librte_vhost/v
Reviewed-by: Rami Rosen
בתאריך יום ג׳, 19 בפבר׳ 2019, 22:38, מאת David Marchand <
david.march...@redhat.com>:
> pthread_getaffinity_np returns a >0 value when failing.
>
> This is mainly for the sake of correctness.
> The only case where it could fail is when passing an incorrect cpuset
> size
The log was printing the device name two times,
first one being supposed to be the driver name.
As we don't know yet the driver name, the log is simplified.
Fixes: 9bf4901d1a11 ("bus/vdev: remove probe with driver name option")
Cc: ferruh.yi...@intel.com
Signed-off-by: Thomas Monjalon
---
drive
These are 2 small fixes.
A bigger cleanup may be needed in vdev:
rte_vdev_init() can be replaced by rte_dev_probe().
Any volunteer? Note: vdev has no maintainer.
Raslan Darawsheh (1):
bus/vdev: fix hotplug twice
Thomas Monjalon (1):
bus/vdev: fix debug message on probing
drivers/bus/vdev/
From: Raslan Darawsheh
In case vdev was already probed, it shouldn't be probed again,
and it should return -EEXIST as error.
There are some checks in vdev_probe() and insert_vdev(),
but a check was missing in vdev_plug().
The check is moved in vdev_probe_all_drivers() which is called
in all code
Hi Yongseok,
Can you let me know how you want to proceed with the below. I think we
could just drop as it's a performance optimization, or maybe you have a
different idea.
thanks,
Kevin.
On 06/02/2019 15:54, Kevin Traynor wrote:
> On 01/14/2019 09:16 PM, Yongseok Koh wrote:
>> On replenishing Rx
Previous ungraceful exit may leave behind un-acked stale
interrupts for slowpath and fastpath.
Interrupt status polling function is started before FLR is
initiated, so we don't have a real way to protect this polling
function invoking an interrupt handler caused due to stale
interrupt status from
On 2/20/19 10:16 AM, Hyong Youb Kim wrote:
On Wed, Feb 20, 2019 at 02:56:36PM +, Radu Nicolau wrote:
On 2/20/2019 12:33 PM, Ferruh Yigit wrote:
On 2/18/2019 3:58 PM, Radu Nicolau wrote:
On 2/18/2019 11:59 AM, Hari Kumar Vemula wrote:
test_alb_reply_from_client test fails due to inco
Reclaim outstanding zmbufs first before freeing memory regions,
otherwise there could be use-after-free.
Fixes: b0a985d1f340 ("vhost: add dequeue zero copy")
Cc: sta...@dpdk.org
Signed-off-by: Tiwei Bie
---
lib/librte_vhost/vhost.h | 6 +
lib/librte_vhost/vhost_user.c | 46 +++
Don't free the zero copy mbufs before they have been consumed,
otherwise there could be use-after-free.
Fixes: b0a985d1f340 ("vhost: add dequeue zero copy")
Cc: sta...@dpdk.org
Signed-off-by: Tiwei Bie
---
lib/librte_vhost/vhost.h | 12
lib/librte_vhost/vhost_user.c | 3 +++
Tiwei Bie (4):
vhost: restore mbuf first when freeing zmbuf
vhost: fix potential use-after-free for zero copy mbuf
vhost: fix potential use-after-free for memory region
doc: improve vhost zero copy guide
doc/guides/prog_guide/vhost_lib.rst | 3 +++
lib/librte_vhost/vhost.h| 3
The mbufs should also be restored in free_zmbufs().
Fixes: b0a985d1f340 ("vhost: add dequeue zero copy")
Fixes: 3ebd930588b7 ("vhost: fix mbuf free")
Cc: sta...@dpdk.org
Signed-off-by: Tiwei Bie
---
lib/librte_vhost/vhost.h | 16
lib/librte_vhost/vhost_user.c | 1 +
lib/l
Highlight that vhost zero copy mbufs should be consumed
as soon as possible.
Signed-off-by: Tiwei Bie
---
doc/guides/prog_guide/vhost_lib.rst | 3 +++
1 file changed, 3 insertions(+)
diff --git a/doc/guides/prog_guide/vhost_lib.rst
b/doc/guides/prog_guide/vhost_lib.rst
index c77df338f..a86c07a
Hi Fiona,
Please see inline.
Thanks,
Anoob
> -Original Message-
> From: Trahe, Fiona
> Sent: Thursday, February 21, 2019 10:33 PM
> To: Anoob Joseph ; Akhil Goyal
> ; Doherty, Declan ; De
> Lara Guarch, Pablo
> Cc: Jerin Jacob Kollanukkaran ; Narayana Prasad Raju
> Athreya ; dev@dpdk.o
Reviewed-by: Rami Rosen
בתאריך יום ה׳, 21 בפבר׳ 2019, 21:01, מאת Thomas Monjalon <
tho...@monjalon.net>:
> The log was printing the device name two times,
> first one being supposed to be the driver name.
> As we don't know yet the driver name, the log is simplified.
>
> Fixes: 9bf4901d1a11 ("b
Use the bitmap lib to help maintain the metrics. we can dynamically
add and remove metrics data. After uninit latency, it can
remove itself from the metrics to make the result from rte_metrics_get_names
more simple.
Signed-off-by: junka
---
lib/librte_latencystats/rte_latencystats.c | 4 +-
Lock-less data structures provide scalability and determinism.
They enable use cases where locking may not be allowed
(for ex: real-time applications).
In the following paras, the term 'memory' refers to memory allocated
by typical APIs like malloc or anything that is representative of
memory, for
Modify the test cases to allocate the QSBR variable dynamically.
Signed-off-by: Honnappa Nagarahalli
---
test/test/test_rcu_qsbr.c | 225 ++---
test/test/test_rcu_qsbr_perf.c | 33 ++---
2 files changed, 144 insertions(+), 114 deletions(-)
diff --git a/test/tes
Keeping the register thread ID size dependent on the max threads
is resulting in performance drops due to address calculations at
run time. Fixing the size of the thread ID registration array
reduces the complexity of address calculation. This change
fixes the maximum number of threads supported to
From: Dharmik Thakkar
Add API positive/negative test cases, functional tests and
performance tests.
Signed-off-by: Malvika Gupta
Signed-off-by: Dharmik Thakkar
Signed-off-by: Honnappa Nagarahalli
Reviewed-by: Gavin Hu
---
test/test/Makefile | 2 +
test/test/autotest_data.py
Add RCU library supporting quiescent state based memory reclamation method.
This library helps identify the quiescent state of the reader threads so
that the writers can free the memory associated with the lock less data
structures.
Signed-off-by: Honnappa Nagarahalli
Reviewed-by: Steve Capper
R
rte_rcu_qsbr_get_memsize API is introduced. This will allow the user
to controll the amount of memory used based on the maximum
number of threads present in the application.
Signed-off-by: Honnappa Nagarahalli
---
lib/librte_rcu/rte_rcu_qsbr.c | 51 ---
lib/librte_rcu/rte_rcu_qsbr.h
> >
> >
> > > > > > > > +/**
> > > > > > > > + * RTE thread Quiescent State structure.
> > > > > > > > + */
> > > > > > > > +struct rte_rcu_qsbr {
> > > > > > > > + uint64_t reg_thread_id[RTE_QSBR_BIT_MAP_ELEMS]
> > > > > > > __rte_cache_aligned;
> > > > > > > > + /**< Registered reade
This patch renames avf to iavf only for the doc files.
Signed-off-by: Leyi Rong
---
MAINTAINERS| 6 +++---
doc/guides/nics/features/{avf.ini => iavf.ini} | 2 +-
.../nics/features/{avf_vec.ini => iavf_vec.ini}| 2 +-
doc/guides/nics/intel_vf.rst
I think the flow list may be accessed in the same time by two
different threads and may cause some errors. Do it need a lock
to protect the flow list?
Thanks
Yunjian
(gdb) bt
#0 0x7f54c9641237 in raise () from /usr/lib64/libc.so.6
#1 0x7f54c9642928 in abort () from /usr/lib64/li
55 matches
Mail list logo
