Add 'uint64_t opaque_data' inside struct rte_security_session.
That allows upper layer to easily associate some user defined
data with the session.
Proposed new layout for:
struct rte_security_session {
void *sess_private_data;
/**< Private session material */
+ uint64_t opaque_data;
+ /**< Opaque user defined data */
};
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com> Acked-by: Mohammad Abdul Awal <mohammad.abdul.awal@intel.com> Acked-by: Fan Zhang <roy.fan.zhang@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
Ian Stokes [Thu, 22 Nov 2018 12:09:09 +0000 (12:09 +0000)]
doc: announce ethdev ABI change for MTU limits info
Maximum and minimum MTU values vary between hardware devices. In
hardware agnostic DPDK applications access to such information would
allow a more accurate way of validating and setting supported MTU values on
a per device basis rather than using a defined default for all devices.
This patch adds a depreciation notice for ``rte_eth_dev_info`` as new
members will be added to represent min and max MTU values. These can be
added to fit a hole in the existing structure for amd64 but not for 32 bit,
as such ABI change will occur as size of the structure will be impacted.
Signed-off-by: Ian Stokes <ian.stokes@intel.com> Acked-by: Ferruh Yigit <ferruh.yigit@intel.com> Acked-by: Thomas Monjalon <thomas@monjalon.net> Acked-by: Andrew Rybchenko <arybchenko@solarflare.com>
Shreyansh Jain [Thu, 15 Nov 2018 13:29:27 +0000 (13:29 +0000)]
doc: announce deprecation of dpaa2 exposed mem structure
rte_dpaa2_memsegs is no more required once the dpaax (pa-va) translation
library has been introduced. This can be made internal (for fallback
operations) in subsequent release.
Signed-off-by: Shreyansh Jain <shreyansh.jain@nxp.com> Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com> Acked-by: Anatoly Burakov <anatoly.burakov@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Thomas Monjalon [Thu, 22 Nov 2018 10:32:23 +0000 (11:32 +0100)]
doc: announce kvargs API change
After processing a kvlist in rte_kvargs_process(),
it may be needed to loop again over kvlist in order to know
whether the key is matched or not.
In order to simplify implementation of kvargs checks,
a new pointer parameter may be used to get the match count.
The change of the function prototype would be as below:
int
rte_kvargs_process(const struct rte_kvargs *kvlist,
const char *key_match,
+ int *match_count,
arg_handler_t handler,
void *opaque_arg)
Signed-off-by: Thomas Monjalon <thomas@monjalon.net> Acked-by: Olivier Matz <olivier.matz@6wind.com> Acked-by: Ferruh Yigit <ferruh.yigit@intel.com> Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Anatoly Burakov [Mon, 19 Nov 2018 13:10:26 +0000 (13:10 +0000)]
doc: document all EAL parameters in one place
Currently, the most complete (but still incomplete) user guide for
EAL command-line parameters resides in user guide for testpmd.
This is wrong on multiple levels, and should not be the case.
To fix it, we have to create a document that lists all supported
EAL command-line arguments. However, because different platforms
support different subsets of available EAL parameters, instead of
creating a single file, we will create a common file in
doc/guides/common containing documentation for EAL parameters
that are supported on all of our supported platforms (Linux and
FreeBSD at the time of this writing).
We will then include this document in the Getting Started guides
for all supported platforms, so that any changes made to
documentation for commonly supported EAL parameters will be
reflected in Getting Started guides for all platforms.
This patch also removes EAL parameters documentation from the
testpmd user guide, and instead adds references to the newly
created documents in both testpmd user guides and in sample
applications guide.
Maxime Coquelin [Thu, 22 Nov 2018 17:09:22 +0000 (18:09 +0100)]
vhost: fix packed ring constants declaration
The packed ring defines were declared only if kernel
header does not declare them.
The problem is that they are not applied in upstream kernel,
and some changes in the names have been required.
This patch declares the defines unconditionally, which
fixes potential build issues.
net/mlx5: fix TPID check for VLAN push action on E-Switch
The VLAN push action on E-Switch supports only 802.1Q (0x8100)
and 802.1AD (0x88A8) Tag Protocol ID (TPID) insertions. The
parameter check for RTE_FLOW_ACTION_TYPE_OF_PUSH_VLAN action
is added.
Fixes: 57123c00c1b8 ("net/mlx5: add Linux TC flower driver for E-Switch flow") Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com> Acked-by: Shahaf Shuler <shahafs@mellanox.com>
net/mlx5: fix VLAN inner ethernet type on E-Switch
The TCA_FLOWER_KEY_VLAN_ETH_TYPE should be specified for the E-Switch
Flows with VLAN and L3 pattern items in the Netlink messages. The patch
adds missing flower key to the messages. This patch partially reverts to
the code smashed by http://patches.dpdk.org/patch/47781
Yanglong Wu [Tue, 20 Nov 2018 05:59:21 +0000 (13:59 +0800)]
net/ixgbe: fix TDH register write
The only time that software should write to the TDH register
is after a reset (hardware reset or CTRL.RST) and
before enabling the transmit function (TXDCTL.ENABLE).
If software were to write to this register while the transmit
function was enabled, the on-chip descriptor buffers might
be invalidated and the hardware could become confused.
Shahed Shaikh [Wed, 21 Nov 2018 21:20:49 +0000 (21:20 +0000)]
net/qede: fix Tx offload mask
Following commit updated the Tx offload mask
commit 1037ed842c37 ("mbuf: fix Tx offload mask").
So, not having PKT_TX_IPV6 and PKT_TX_IPV4 in qede PMDs supported
Tx offload mask breaks TSO support since application will fail in transmit
prepare function.
Rafal Kozik [Wed, 21 Nov 2018 08:21:14 +0000 (09:21 +0100)]
net/ena: fix out of order completion
rx_buffer_info should be refill not linearly, but out of order.
IDs should be taken from empty_rx_reqs array.
rx_refill_buffer is introduced to temporary storage
bulk of mbufs taken from pool.
In case of error unused mbufs are put back to pool.
Fixes: c2034976673d ("net/ena: add Rx out of order completion") Cc: stable@dpdk.org Signed-off-by: Rafal Kozik <rk@semihalf.com> Acked-by: Michal Krawczyk <mk@semihalf.com>
Jasvinder Singh [Tue, 20 Nov 2018 13:00:15 +0000 (13:00 +0000)]
net/softnic: fix undefined device info fields
The memcpy operation overwrites the device info fields set
by ethdev API and leaves the device field undefined.
Thus, replaces memcpy by only updating the required fields
at the driver layer.
error log: testpmd> show port info 1
**** Infos for port 1 ****
MAC address: 00:00:00:00:00:00
Device name: net_softnic0
Segmentation fault.
port_infos_display (port_id=1) at /dpdk/app/test-pmd/config.c:418
if (dev_info.device->devargs && dev_info.device->devargs->args)
Fan Zhang [Mon, 19 Nov 2018 09:38:31 +0000 (09:38 +0000)]
doc: fix typo in vhost-crypto app guide
This patch fixes the mismatch keyword in vhost_crypto sample
application guide.
Fixes: 709521f4c2cd ("examples/vhost_crypto: support multi-core") Cc: stable@dpdk.org Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> Acked-by: Marko Kovacevic <marko.kovacevic@intel.com>
Joyce Kong [Thu, 22 Nov 2018 07:37:07 +0000 (15:37 +0800)]
config: enable more than 128 cores for arm64
When running dpdk applications on cores whose ids are bigger
than original max_core setting, eal error as below:
EAL: Detected 104 lcore(s)
EAL: Detected 2 NUMA nodes
EAL: invalid core list
-l CORELIST core indexes between 0 and 128
The fix is to increase max_core to 256 on arm64 platform.
Fixes: b3ce00e5 ("mk: introduce ARMv8 architecture") Cc: stable@dpdk.org Signed-off-by: Joyce Kong <joyce.kong@arm.com> Reviewed-by: Gavin Hu <gavin.hu@arm.com> Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Thomas Monjalon [Fri, 16 Nov 2018 16:58:54 +0000 (17:58 +0100)]
app/eventdev: fix possible uninitialized variable
When compiling with -O1, this error can appear:
app/test-eventdev/test_pipeline_common.c:332:6: error:
‘ret’ may be used uninitialized in this function
If there is no device, then ret is returned without being initialized.
It is fixed by setting 0 as initial value.
Fixes: 032a965a8f1d ("app/eventdev: support Tx adapter") Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Thomas Monjalon [Fri, 16 Nov 2018 16:58:53 +0000 (17:58 +0100)]
eventdev: fix possible uninitialized variable
When compiling with -O1, this error can appear:
lib/librte_eventdev/rte_event_eth_tx_adapter.c:705:6: error:
‘ret’ may be used uninitialized in this function
If tx_queue_id is -1 and nb_queues is 0, then ret is returned
without being initialized.
It is fixed by setting 0 as initial value.
Fixes: a3bbf2e09756 ("eventdev: add eth Tx adapter implementation") Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Thomas Monjalon [Fri, 16 Nov 2018 16:58:52 +0000 (17:58 +0100)]
net/mlx4: fix possible uninitialized variable
When compiling with gcc -O1, this error appears:
drivers/net/mlx4/mlx4_ethdev.c: In function ‘mlx4_rxmode_toggle’:
rte_log.h:321:3: error:
‘mode’ may be used uninitialized in this function
The function mlx4_rxmode_toggle is never called with a value which
is not in the switch block, but GCC complains about it with -O1.
So the default case is "fixed" by setting string "undefined".
Fixes: eacaac7bae36 ("net/mlx4: restore promisc and allmulti support") Cc: stable@dpdk.org Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Thomas Monjalon [Fri, 16 Nov 2018 16:58:51 +0000 (17:58 +0100)]
kni: fix possible uninitialized variable
This error can be raised:
lib/librte_kni/rte_kni.c:531:15: error:
'req' may be used uninitialized in this function
It should not happen because kni_fifo_get() would return 0 if
req is not initialized, so the function would return before using req.
But GCC complains about it in -O1 optimization,
and a NULL initialization is harmless here.
Thomas Monjalon [Fri, 16 Nov 2018 16:58:50 +0000 (17:58 +0100)]
eal: fix build with -O1
In case of optimized compilation, RTE_BUILD_BUG_ON use an external
variable which is neither defined, nor used.
It seems not optimized out in case of OPDL compiled with clang -O1:
opdl_ring.c: undefined reference to `RTE_BUILD_BUG_ON_detected_error'
clang-6.0: fatal error: linker command failed with exit code 1
Fixes: af75078fece3 ("first public release") Cc: stable@dpdk.org Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Anatoly Burakov [Wed, 21 Nov 2018 12:05:03 +0000 (12:05 +0000)]
eal: deprecate and rename bsf64 function
Rename rte_bsf64 to rte_bsf64_safe (this is a "safe" version in
that it prevents undefined behavior by checking if incoming
parameter is zero) and move it to common header.
Anatoly Burakov [Wed, 21 Nov 2018 12:05:02 +0000 (12:05 +0000)]
eal: remove useless code in bsf64 function
RTE_BITMAP_OPTIMIZATIONS was never set to 0 and makes no sense
anyway, so remove all code related to it. Also, drop the "likely"
for bsf64 code, because it's a generic function and we cannot
make any assumptions about likely values of incoming arguments.
Anatoly Burakov [Tue, 20 Nov 2018 16:18:46 +0000 (16:18 +0000)]
ipc: fix access after async request failure
Previous fix for rte_panic has moved setting of alarm before
sending the message. This means that whether we send a message,
the alarm would still trigger. The comment noted that cleanup
would happen in the alarm handler, but that's not what actually
happened - instead, in the event of failed send we freed the
memory in-place, before putting the request on the queue.
This works OK when the message is sent, but when sending the
message fails, the alarm would still trigger with a pointer
argument that points to non-existent memory, and cause
memory corruption.
There probably is a "proper" fix for this issue, with correct
handling of sent vs. unsent requests, however it would be
simpler just to sacrifice the sent request in the (extremely
unlikely) event of alarm set failing. The other process would
still send a response, but it will be ignored by the sender.
Anatoly Burakov [Mon, 19 Nov 2018 10:33:12 +0000 (10:33 +0000)]
usertools: skip empty categories in devices status
If there aren't any devices of a particular category on user's
system, we still display them, which is bad for usability. Fix
devbind to not print out a category unless there are devices in
it.
Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com> Reviewed-by: David Hunt <david.hunt@intel.com>
David Hunt [Fri, 16 Nov 2018 15:47:21 +0000 (15:47 +0000)]
doc: add jansson link instructions for power app
Add more info to the docs on how to set environmental variable to
correctly build the vm_power_manager sample app in a cross compile
or multilib environment by setting PKG_CONFIG_LIBDIR.
Signed-off-by: David Hunt <david.hunt@intel.com> Acked-by: Marko Kovacevic <marko.kovacevic@intel.com>
Tone Zhang [Wed, 26 Sep 2018 05:08:47 +0000 (13:08 +0800)]
doc: clarify IOMMU disabling for uio_pci_generic
If the devices used for DPDK are bound to the "uio_pci_generic" kernel
module, the IOMMU should be disabled in order not to break the IO
transmission because of the virtual / physical address mapping.
The patch clarifies the IOMMU configurations on both x86_64 and arm64
systems.
Signed-off-by: Tone Zhang <tone.zhang@arm.com> Reviewed-by: Honnappa Nagarahalli <honnappa.nagarahalli@arm.com> Reviewed-by: Stephen Hemminger <stephen@networkplumber.org> Reviewed-by: Rami Rosen <roszenrami@gmail.com> Acked-by: Gavin Hu <gavin.hu@arm.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Luca Boccassi <bluca@debian.org>
Kevin Traynor [Wed, 10 Oct 2018 14:14:16 +0000 (15:14 +0100)]
doc: note minimun Linux version increase for 19.02
Updating docs to reflect decision made at the techboard
that the min kernel version should be bumped from 3.2 to
the latest longterm stable release (3.16), but that
compatibility for commonly used distribution kernels should
be kept also.
Signed-off-by: Kevin Traynor <ktraynor@redhat.com> Acked-by: Maxime Coquelin <maxime.coquelin@redhat.com> Acked-by: Thomas Monjalon <thomas@monjalon.net>
Ferruh Yigit [Mon, 5 Nov 2018 17:28:44 +0000 (17:28 +0000)]
doc: add KNI behaviour change in release notes
Commit 89397a01ce4a ("kni: set default carrier state of interface")
changes the KNI interface default carrier status. Which prevents traffic
flow by default and may break some existing usage / testing.
Document this behavior change in release notes.
Fixes: c6fd54f28c24 ("kni: add function to set link state on kernel interface") Fixes: 89397a01ce4a ("kni: set default carrier state of interface") Fixes: 724beb913b44 ("examples/kni: monitor and update link state continually") Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Dan Gora [Mon, 5 Nov 2018 19:54:13 +0000 (17:54 -0200)]
doc: add new KNI features to release notes
Add the new module parameter for the KNI kernel module, the new command
line flag for the KNI sample application, and the new API function
'rte_kni_update_link()' to the release note.
Fixes: c6fd54f28c24 ("kni: add function to set link state on kernel interface") Fixes: 89397a01ce4a ("kni: set default carrier state of interface") Fixes: 724beb913b44 ("examples/kni: monitor and update link state continually") Signed-off-by: Dan Gora <dg@adax.com> Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Neil Horman [Thu, 1 Nov 2018 13:54:10 +0000 (09:54 -0400)]
devtools: fix regex in symbol addition check
The regex to determine the end of the map file chunk in a patch seems to
be wrong, It was using perl regex syntax, which awk doesn't appear to
support (I'm still not sure how it was working previously). Regardless,
it wasn't triggering and as a result symbols were getting added to the
mapdb that shouldn't be there.
Fix it by converting the regex to use traditional posix syntax, matching
only on the negation of the character class [^map]
Tested and shown to be working on the ip_frag patch set provided by
doucette@bu.edu
Fixes: 4bec48184e33 ("devtools: add checks for ABI symbol addition") Cc: stable@dpdk.org Reported-by: Cody Doucette <doucette@bu.edu> Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Phil Yang [Fri, 19 Oct 2018 11:00:38 +0000 (19:00 +0800)]
test: release ring resources after PMD perf test
Need to release the port and the ring resources after test. Otherwise,
it will cause failure to allocate memory when reentry the test.
Fixes: ea764af ("app/test: add performance test for ring driver") Cc: stable@dpdk.org Signed-off-by: Phil Yang <phil.yang@arm.com> Reviewed-by: Gavin Hu <gavin.hu@arm.com>
Anatoly Burakov [Thu, 15 Nov 2018 12:18:05 +0000 (12:18 +0000)]
test: fix hugepage file handling in EAL flags autotest
Before 18.05, DPDK could not release memory back to the system
neither at runtime nor before shutting down. Over the course of
18.05 up to 18.11, code was introduced to release memory at
runtime, as well as an rte_eal_cleanup() function that is supposed
to release all EAL-allocated memory before shutting down DPDK.
When 3f9e31d71d63 ("test: clean up on exit") was introduced, the
test application started to use rte_eal_cleanup() to release all
used memory after execution. However, the EAL flags autotest
still relies on the old behavior of leaving stuff behind in the
hugetlbfs.
The fix is twofold. First, the test to check for leftover files
in hugetlbfs is no longer valid as it is, because test application
now removes all files from hugetlbfs after exit. However, if we
use the --legacy-mem option, then old behavior of leaving files
in hugetlbfs after execution is restored. So the first fix is to
add --legacy-mem to all the tests that expect files in hugetlbfs
to be leftover.
However, we also need to test if default memory mode *doesn't*
leave any files behind, so we also extend the test to check for
these scenarios as well. So, both memtest1 and memtest2 are run
in legacy and default mem modes, and are checked for any leftover
files that are or are not supposed to be there.
Fixes: 3f9e31d71d63 ("test: clean up on exit") Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
examples/ipv4_multicast: fix leak of cloned packets
The ipv4_multicast sample application was dropping packets
when using mbuf clone. When creating an L2 header and copying
metadata from the source packet, the ol_flags were also copied
along with all the other metadata. Because the cloned packet
had IND_ATTACHED_MBUF flag set in its ol_flags,
this caused the packets to never be freed when using rte_pktmbuf_free.
Since copying ol_flags from the cloned packet is
not necessary in the first place, just don't do it.
Fixes: af75078fece3 ("first public release") Cc: stable@dpdk.org Reported-by: Dong Wang <dong1.wang@intel.com> Signed-off-by: Herakliusz Lipiec <herakliusz.lipiec@intel.com> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com> Acked-by: Dong Wang <dong1.wang@intel.com>
Qi Zhang [Thu, 15 Nov 2018 01:37:42 +0000 (09:37 +0800)]
net/pcap: enable data path for secondary process
Private vdev was the way previously, when pdump developed, now with
shared device mode on virtual devices, pcap data path in secondary
is not working.
When secondary adds a virtual device, related data transferred to
primary and primary creates the device and shares device back with
secondary. When pcap device created in primary, pcap handlers
(pointers) are process local and they are not valid for secondary
process. This breaks secondary.
So we can't directly share the pcap handlers, but need to create a new
set of handlers for secondary, that's what we done in this patch.
Qi Zhang [Thu, 15 Nov 2018 01:37:41 +0000 (09:37 +0800)]
net/pcap: move handler to process private
This is prework for data path enabling for secondary process.
To prevent pcap handler opened by one process be overwritten by
another process, each process should have their private copy,
`rte_eth_dev->process_private` is exactly what we needed.
Zhirun Yan [Fri, 16 Nov 2018 15:25:00 +0000 (15:25 +0000)]
net/avf: fix Tx offload mask
Tx offload mask is updated in following commit 1037ed842c37
("mbuf: fix Tx offload mask"). Currently, the new added offload
flags are not supported in PMD and application will fail to call
PMD transmit prepare function.
Fan Zhang [Wed, 14 Nov 2018 11:16:42 +0000 (11:16 +0000)]
examples/vhost_crypto: fix zero copy
This patch fixes the zero copy enable problem for vhost crypto
sample application.
For some Crypto PMDs such as AESNI-MB and AESNI-GCM the data to
be processed will be made a copy in the same buffer but next to the
data. For example, to encrypt 64 bytes data the PMD will copy this
data from offset 64 to offset 123. This requires the application
provides the buffer with at least double of the data size.
However there is no way for VMs to know this limitation. When
zero-copy is enabled in Vhost the PMD may overwrite the buffer
next to the VM data to be processed, and further cause problems
such as Segmentation Fault or even worse, crashes the VM.
To fix the problem the user should avoid enabling the zero copy
for these Crypto PMDs. This patch adds the checking of the PMD
names to see if zero copy can be applied.
Fixes: 709521f4c2cd ("examples/vhost_crypto: support multi-core") Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Anoob Joseph [Wed, 14 Nov 2018 09:22:03 +0000 (09:22 +0000)]
net/octeontx: fix failures when available ports > queues
When eth_octeontx is initialized before event_octeontx, eth_octeontx
would initiate setting up of the event device with the max ports and
queues available. If number of ports is more than the number of queues,
some ports would be left unused when the ports and queues are mapped
1:1. But even in that case the ports need to be setup, or otherwise it
would lead to a segmentation fault when event device is started.
Fixes: f7be70e5130e ("net/octeontx: add net device probe and remove") Cc: stable@dpdk.org Signed-off-by: Anoob Joseph <anoob.joseph@caviumnetworks.com> Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com> Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Dekel Peled [Thu, 15 Nov 2018 15:17:13 +0000 (17:17 +0200)]
net/mlx5: fix MPLS item validation
Update the mlx5_flow_validate_item_mpls() function to allow
MPLS over IP, UDP, and GRE.
Modify the flow_dv_validate() function with the new logic introduced
in previous patch of this series: set new variable last_item
after each validation, update item_flags after each item iteration.
The new variable last_item is sent to mlx5_flow_validate_item_mpls()
and used to validate the MPLS item.
Same change implemented in flow_verbs_validate().
Update the mlx5_flow_validate_item_mpls() function to verify that
device configuration has mpls_en set to 1.
This code was added earlier this year, but unintentionaly removed
in recent versions.
Dekel Peled [Thu, 15 Nov 2018 15:17:12 +0000 (17:17 +0200)]
net/mlx5: add MPLS to Direct Verbs flow engine
The support in MPLS on this flow engine was overlooked. It's absence is
critical because there are required actions for MPLS which can be done
only with the DV engine.
To set correctly the MPLS filter, we need to reason about the flow item
before the MPLS (UDP, GRE or other).
To do that, a new variable last_item was added and updated after each
translation. the full item flags are updated after each item iteration.
Yongseok Koh [Thu, 15 Nov 2018 10:29:19 +0000 (10:29 +0000)]
net/mlx5: optimize Tx doorbell write
Unnecessary volatile attribute keeps compiler from further optimizing the
code and this results in a little performance drop (~2%). Because of memory
barriers, it is safe to remove.
There's some performance drop due to extra condition checks on the
datapath. Checking for external memory registration should be consolidated
to the existing bottom-half.
There's some performance drop due to extra condition checks on the
datapath. Checking for external memory registration should be consolidated
to the existing bottom-half.
The flow_dv_query() just returns -ENOTSUP value and does not
set provided error parameter structure, that crashes the
port_flow_query(). The patch fixes flow_db_query(), now it
sets an error parameter structure.
net/mlx5: add E-switch rule hardware offload flag check
This patch adds the in_hw flag check for tc flower filter rules.
If rule is applied without skip_sw flag set driver should check
whether the in_hw flag is set after rule applying. If no in_hw
flag set found it means the rule is not hardware offloaded and
error should be returned to application.
net/mlx5: prepare to add E-switch rule flags check
The tc flower filter rules are used to control E-switch from
the application side. In order to gain garanteed rule hardware
offload the skip_sw flag should be specified while applying the
rule. But some tc rules is rejected by kernel if skip_sw flag
is set by design. Currently this regards VXLAN tunneling rules,
which are applied to special VXLAN virtual devices. Albeit these
rules are applied with skip_sw flag reset kernel tries to
perform hardware offload. If kernel succeeded the in_hw flag
is set in rule properties and application should check this
flag to get know whether the applied rule is actually hardware
offloaded.
This patch prepares for the rule flags query and check in_hw flag.
The driver checks only the rules with skip_sw flag reset, so we
need to test the actual flags of rule is being applied. The pointer
to flags field into translated rule is introduced and flags are
tested directly in the rule body. It is more reliable than save and
check flags copy.
Also patch swaps the flow_tcf_apply() and flow_tcf_remove(),
we are going to call flow_tcf_remove() from flow_tcf_appy() if
no in_hw flag set found in applied rule.
While receiving the Netlink reply messages we should stop at DONE
or ACK message. The existing implementation stops at DONE message
or if no multiple message flag set ( NLM_F_MULTI). It prevents
the single query requests from working, these requests send the
single reply message without multi-message flag followed by
ACK message. This patch fixes receiving part of Netlink
communication routine.
This patch removes the unused message length parameter, we
do not send multiple commands in the single message anymore,
message length can be taken from the prepared message header,
so length parameter can be removed.
Xiaolong Ye [Thu, 15 Nov 2018 07:20:05 +0000 (15:20 +0800)]
net/ixgbe: fix maximum wait time in comment
As the code has changed the max wait time to 1000ms, the comment should
be changed accordingly.
Fixes: 64f1c8539c8c ("net/ixgbe: wait longer for link after fiber MAC setup") Cc: stable@dpdk.org Signed-off-by: Xiaolong Ye <xiaolong.ye@intel.com> Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Beilei Xing [Thu, 15 Nov 2018 03:17:02 +0000 (11:17 +0800)]
net/i40e: fix X710 Rx after reading some registers
There's an issue that X710 can't receive any packet after
reading some special registers. That's because these
registers are only valid for X722, read access for non-X722
will cause ECC error.
Jeff Guo [Thu, 15 Nov 2018 09:18:24 +0000 (17:18 +0800)]
app/testpmd: workaround deadlock in hot-unplug callback
Because the user's callback is invoked in eal interrupt callback, the
interrupt callback need to be finished before it can be unregistered
when detaching device. So finish callback soon and use a deferred
removal to detach device is need.
It is a workaround, once the device detaching be moved into the eal in
the future, the deferred removal could be deleted. This patch aim to
add this workaround and refine the function name and the description to
be more explicit and comment the limitation.
Fixes: 2049c5113fe8 ("app/testpmd: use hotplug failure handler") Signed-off-by: Jeff Guo <jia.guo@intel.com>
Jeff Guo [Thu, 15 Nov 2018 09:18:23 +0000 (17:18 +0800)]
vfio: add lock for hot-unplug failure handler
When the sigbus handler be enabled for hot-unplug, whatever hot-unplug
sigbus or origin sigbus occur, the sigbus handler will be invoked and
it will access the bus and device. While in the control path, the vfio
req handler also will process the bus and device, so a protection of
the resources in vfio req handler should be need. This patch add a lock
in vfio req handler when process bus and device resource, to avoid the
synchronization issue when device be hot-unplugged.
git.droids-corp.org / dpdk.git / log