Shweta Choudaha [Thu, 16 May 2019 22:03:31 +0000 (23:03 +0100)]
net/igb: fix PHY status if PHY reset is not blocked
When PHY reset is blocked as is the case when BMC is connected via NC-SI
do not set GO_LINKD bit in PHY power management register in dev_stop as
this will disconnect the PHY. Also, in dev_close clear the GO_LINKD
bit only if PHY reset is not blocked
Fixes: 3af34dec0b41 ("igb: force phy power up/down") Cc: stable@dpdk.org Signed-off-by: Shweta Choudaha <shweta.choudaha@att.com> Reviewed-by: Xiaolong Ye <xiaolong.ye@intel.com>
Ying A Wang [Thu, 24 Oct 2019 21:39:59 +0000 (05:39 +0800)]
net/ice: fix packet type table
The original packet type table's hw-ptype index to sw-ptype
mapping is wrong. This patch fixed the issue.
Fixes: ab9dccf09af9 ("net/ice: support packet type getting") Cc: stable@dpdk.org Signed-off-by: Ying A Wang <ying.a.wang@intel.com> Reviewed-by: Xiaolong Ye <xiaolong.ye@intel.com>
drivers: process shared link dependencies as for libs
For the public APIs of DPDK libraries we run checks for correct use of
experimental tags, and also do dynamic generation of the version file to
its window's equivalent. Although must drivers don't export APIs, some do,
so these checks are relevant and should be copied from lib/meson.build to
drivers/meson.build.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Luca Boccassi <bluca@debian.org>
buildtools: remove make-specific from symbols check
The check-experimental-syms.sh script was finding the map-list-symbol.sh
script using $RTE_SDK, which is the variable set when using the "make"
build system. To make this script more independent, we just use the current
path of the script as the location to find its companion script.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Luca Boccassi <bluca@debian.org>
crypto/openssl: use local copy for session contexts
Session contexts are used for temporary storage when processing a
packet.
If packets for the same session are to be processed simultaneously on
multiple cores, separate contexts must be used.
Note: with openssl 1.1.1 EVP_CIPHER_CTX can no longer be defined as a
variable on the stack: it must be allocated. This in turn reduces the
performance.
examples/ipsec-secgw: fix library mode selection in tests
Right now majority of lib mode functional tests enable
library mode with '-w N' option.
It worked till recently, as legacy mode didn't support replay window.
Now it changed.
To fix - use '-l' option to enable library mode explicitly.
Fixes: 929784452094 ("examples/ipsec-secgw: add scripts for functional test") Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Hemant Agrawal [Wed, 6 Nov 2019 06:54:13 +0000 (12:24 +0530)]
ipsec: remove redundant replay window size
The rte_security lib has introduced replay_win_sz,
so it can be removed from the rte_ipsec lib.
The relevant tests, app are also update to reflect
the usages.
Note that esn and anti-replay fileds were earlier used
only for ipsec library, they were enabling the libipsec
by default. With this change esn and anti-replay setting
will not automatically enabled libipsec.
Hemant Agrawal [Wed, 6 Nov 2019 06:54:12 +0000 (12:24 +0530)]
security: add anti replay window size
At present the ipsec xfrom is missing the important step
to configure the anti replay window size.
The newly added field will also help in to enable or disable
the anti replay checking, if available in offload by means
of non-zero or zero value.
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com> Acked-by: Anoob Joseph <anoobj@marvell.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
As per RFC4868, SHA-256 should use 128 bits of ICV. Fixes: b5350285ce6e ("examples/ipsec-secgw: support SHA256 HMAC") Cc: stable@dpdk.org Signed-off-by: Vakul Garg <vakul.garg@nxp.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
Vakul Garg [Wed, 6 Nov 2019 05:17:30 +0000 (10:47 +0530)]
crypto/dpaax_sec: use AES-CTR initial counter as 1
As per RFC3686, the initial aes-ctr counter value should be '1' for use
in ipsec. The patches changes SEC descriptor for using correct counter
value. In addition, it drops a redundant parameter for passing IV while
creating the descriptor.
Pablo de Lara [Fri, 1 Nov 2019 13:39:23 +0000 (13:39 +0000)]
crypto/aesni_gcm: support in-place chained mbufs
IPSec Multi buffer library supports encryption on multiple segments.
When dealing with chained buffers (multiple segments), as long as
the operation is in-place, the destination buffer does not have to
be contiguous (unlike in the case of out-of-place operation).
Therefore, the limitation of not supporting in-place chained mbuf
can be removed.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
The example IPsec application does not work properly when using
AES-GCM with crypto_openssl.
ESP with AES-GCM uses standard 96bit long algorithm IV ([1]) which
later concatenated with be32(1) forms a J0 block. GCM specification
([2], chapter 7.1) states that when length of IV is different than
96b, in order to format a J0 block, GHASH function must be used.
According to specification ([2], chapter 5.1.1) GCM implementations
should support standard 96bit IVs, other lengths are optional. Every
DPDK cryptodev supports 96bit IV and few of them supports 128bit
IV as well (openssl, mrvl, ccp). When passing iv::length=16 to a
cryptodev which does support standard IVs only (e.g. qat) it
implicitly uses starting 96 bits. On the other hand, openssl follows
specification and uses GHASH to compute J0 for that case which results
in different than expected J0 values used for encryption/decryption.
Fix an inability to use AES-GCM with crypto_openssl by changing IV
length to the standard value of 12.
Akhil Goyal [Tue, 5 Nov 2019 21:07:12 +0000 (02:37 +0530)]
crypto/dpaa2_sec: fix length retrieved from hardware
FD retrieved from SEC after crypto processing provides
an updated length of the buffer which need to be updated
in mbuf. The difference in length can be negative hence
changing diff to int32_t from uint32_t.
Coverity reported about two division by zero:
*** CID 350344: Incorrect expression (DIVIDE_BY_ZERO)
Although in fact these dividers will never be equal to 0,
adding explicit checks in lookup() to make coverity happy
will not affect the execution speed.
Fixes: 908be0651a5a ("app/test-sad: add test application for IPsec SAD") Signed-off-by: Vladimir Medvedkin <vladimir.medvedkin@intel.com> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Add tests for offload fallback feature; add inbound config modificator
SGW_CFG_XPRM_IN (offload fallback setting can be set only for inbound
SAs). Tests are using cryptodev for outbound SA.
To test fragmentation with QAT set:
MULTI_SEG_TEST="--reassemble=4096 --cryptodev_mask=0x5555"
Signed-off-by: Marcin Smoczynski <marcinx.smoczynski@intel.com> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Tested-by: Bernard Iremonger <bernard.iremonger@intel.com>
Due to fragment loss on highly saturated links and long fragment
lifetime, ipsec-secgw application quickly runs out of free reassembly
buckets. As a result new fragments are being dropped.
Introduce --frag-ttl option which allow user to lower default fragment
lifetime which solves problem of saturated reassembly buckets with high
bandwidth fragmented traffic.
Signed-off-by: Marcin Smoczynski <marcinx.smoczynski@intel.com> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Tested-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Inline processing is limited to a specified subset of traffic. It is
often unable to handle more complicated situations, such as fragmented
traffic. When using inline processing such traffic is dropped.
Introduce fallback session for inline crypto processing allowing
handling packets that normally would be dropped. A fallback session is
configured by adding 'fallback' keyword with 'lookaside-none' parameter
to an SA configuration. Only 'inline-crypto-offload" as a primary
session and 'lookaside-none' as a fall-back session combination is
supported by this patch.
Fallback session feature is not available in the legacy mode.
Signed-off-by: Marcin Smoczynski <marcinx.smoczynski@intel.com> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Tested-by: Bernard Iremonger <bernard.iremonger@intel.com>
Cleanup ipsec_sa structure by removing every field that is already in
the rte_ipsec_session structure:
* cryptodev/security session union
* action type
* offload flags
* security context
References to abovementioned fields are changed to direct references
to matching fields of rte_ipsec_session structure.
Such refactoring is needed to introduce many sessions per SA feature,
e.g. fallback session for inline offload processing.
Signed-off-by: Marcin Smoczynski <marcinx.smoczynski@intel.com> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com> Acked-by: Radu Nicolau <radu.nicolau@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Tested-by: Bernard Iremonger <bernard.iremonger@intel.com>
David Marchand [Fri, 8 Nov 2019 18:21:14 +0000 (19:21 +0100)]
ci: remove LTO job
The LTO job using gcc-7 has two issues at the moment:
- warnings about implicit fallthroughs trigger build errors:
In file included from ...common/include/rte_memory.h:22:0,
from ...linux/eal/eal_hugepage_info.c:24:
...common/include/rte_common.h: In function ‘rte_str_to_size’:
...common/include/rte_common.h:744:27: error: this statement may
fall through [-Werror=implicit-fallthrough=]
case 'G': case 'g': size *= 1024; /* fall-through */
~~~~~^~~~~~~
- if we disable this warning, linking the binaries takes too much time
and the job is terminated by Travis because it reaches the maximum
time limit
Fixes: 098cc0fea3be ("build: add option to enable LTO") Signed-off-by: David Marchand <david.marchand@redhat.com> Acked-by: Thomas Monjalon <thomas@monjalon.net>
Ali Alnubani [Mon, 4 Nov 2019 13:30:08 +0000 (15:30 +0200)]
mk: fix build on arm64
OcteonTx was disabled for causing an internal compiler error on old gcc
versions.
See commit 4f760550a093 ("mk: disable OcteonTx for buggy compilers").
The condition that was added later to apply disabling OcteonTx
only on arm64 caused the condition to never be obeyed because it
compares ["arm64"] to [arm64].
This fixes the condition by using RTE_ARCH instead of CONFIG_RTE_ARCH,
because the former has the quotes removed, while the later doesn't.
Fixes: f3af3e44a444 ("mk: disable OcteonTx for buggy compilers only on arm64") Cc: stable@dpdk.org Signed-off-by: Ali Alnubani <alialnu@mellanox.com> Acked-by: Thomas Monjalon <thomas@monjalon.net> Acked-by: Jerin Jacob <jerinj@marvell.com>
Ali Alnubani [Wed, 23 Oct 2019 11:56:09 +0000 (14:56 +0300)]
doc: disable dot graph in doxygen
Ubuntu ships with a patched version of doxygen that enables
HAVE_DOT (disabled by default). Enabling this option causes the warning:
"""
warning: Included by graph for 'rte_common.h' not generated,
too many nodes. Consider increasing DOT_GRAPH_MAX_NODES
"""
This reproduces with doxygen version 1.8.13 and
dot - graphviz version 2.40.1 on Ubuntu 18.04.
This will force doxygen not to assume that dot (part of Graphviz)
is installed, and will result in dot not being used for visualization.
If someone still needs to generate the graphs, the following can
be considered:
- Increase DOT_GRAPH_MAX_NODES to a large value.
- Set HAVE_DOT for more powerful graphs.
- Set DOT_IMAGE_FORMAT=svg to generate svg images.
- Set INTERACTIVE_SVG=YES to allow zooming and panning.
David Marchand [Fri, 8 Nov 2019 10:04:42 +0000 (11:04 +0100)]
examples/l2fwd-event: fix build on RHEL 7.6
CC l2fwd_event_generic.o
.../l2fwd_event_generic.c: In function
‘l2fwd_rx_tx_adapter_setup_generic’:
.../l2fwd_event_generic.c:203:3: error: missing initializer for field
‘impl_opaque’ of ‘struct <anonymous>’
[-Werror=missing-field-initializers]
}
^
In file included from .../l2fwd_event_generic.c:10:0:
.../include/rte_eventdev.h:1057:12: note: ‘impl_opaque’ declared here
uint8_t impl_opaque;
^
CC l2fwd_event_internal_port.o
.../l2fwd_event_internal_port.c: In function
‘l2fwd_rx_tx_adapter_setup_internal_port’:
.../l2fwd_event_internal_port.c:201:3: error: missing initializer for
field ‘impl_opaque’ of ‘struct <anonymous>’
[-Werror=missing-field-initializers]
}
^
In file included from .../l2fwd_event_internal_port.c:10:0:
.../include/rte_eventdev.h:1057:12: note: ‘impl_opaque’ declared here
uint8_t impl_opaque;
^
Fixes: 50f05aa6ed9a ("examples/l2fwd-event: setup Rx/Tx adapter") Signed-off-by: David Marchand <david.marchand@redhat.com> Acked-by: Kevin Traynor <ktraynor@redhat.com>
Kevin Traynor [Wed, 6 Nov 2019 19:01:59 +0000 (19:01 +0000)]
net/ipn3ke: remove useless if statement
Coverity complains that this statement is not needed as the goto
label is on the next line anyway. Remove the if statement.
653 ret = ipn3ke_cfg_parse_i40e_pf_ethdev(afu_name, pf_name);
CID 337930 (#1 of 1): Identical code for different branches
(IDENTICAL_BRANCHES)identical_branches: The same code is executed
when the condition ret is true or false, because the code in the
if-then branch and after the if statement is identical. Should
the if statement be removed?
654 if (ret)
655 goto end;
implicit_else: The code from the above if-then branch is identical
to the code after the if statement.
656end:
Coverity issue: 337930 Fixes: c01c748e4ae6 ("net/ipn3ke: add new driver") Cc: stable@dpdk.org Signed-off-by: Kevin Traynor <ktraynor@redhat.com> Reviewed-by: David Marchand <david.marchand@redhat.com> Reviewed-by: Rosen Xu <rosen.xu@intel.com>
Kevin Traynor [Wed, 6 Nov 2019 19:01:58 +0000 (19:01 +0000)]
net/ipn3ke: fix TM commit check
Coverity is complaining about identical code regardless of which branch
of the if else is taken. Functionally it means an error will always be
returned if this if else is hit. Remove the else branch.
CID 337928 (#1 of 1): Identical code for different branches
(IDENTICAL_BRANCHES)identical_branches: The same code is executed
regardless of whether n->level != IPN3KE_TM_NODE_LEVEL_COS ||
n->n_children != 0U is true, because the 'then' and 'else' branches
are identical. Should one of the branches be modified, or the entire
'if' statement replaced?
1506 if (n->level != IPN3KE_TM_NODE_LEVEL_COS ||
1507 n->n_children != 0) {
1508 return -rte_tm_error_set(error,
1509 EINVAL,
1510 RTE_TM_ERROR_TYPE_UNSPECIFIED,
1511 NULL,
1512 rte_strerror(EINVAL));
else_branch: The else branch, identical to the then branch.
1513 } else {
1514 return -rte_tm_error_set(error,
1515 EINVAL,
1516 RTE_TM_ERROR_TYPE_UNSPECIFIED,
1517 NULL,
1518 rte_strerror(EINVAL));
1519 }
Coverity issue: 337928 Fixes: c820468ac99c ("net/ipn3ke: support TM") Cc: stable@dpdk.org Signed-off-by: Kevin Traynor <ktraynor@redhat.com> Reviewed-by: Rosen Xu <rosen.xu@intel.com>
Kevin Traynor [Wed, 6 Nov 2019 19:01:57 +0000 (19:01 +0000)]
common/cpt: fix possible null dereference
Coverity complains that ctrl_flags is set to NULL at the start
of the function and it may not have been set before there is a
jump to fc_success and it is dereferenced.
Kevin Traynor [Wed, 6 Nov 2019 19:01:56 +0000 (19:01 +0000)]
net/pcap: fix argument checks
Previously rx/tx_queues were passed into eth_from_pcaps_common()
as ptrs and were checked for being NULL.
In commit da6ba28f0540 ("net/pcap: use a struct to pass user options")
that changed to pass in a ptr to a pmd_devargs_all which contains
the rx/tx_queues.
The parameter checking was not updated as part of that commit and
coverity caught that there was still a check if rx/tx_queues were
NULL, apparently after they had been dereferenced.
In fact as they are a members of the devargs_all struct, they will
not be NULL so remove those checks.
1231 struct pmd_devargs *rx_queues = &devargs_all->rx_queues;
1232 struct pmd_devargs *tx_queues = &devargs_all->tx_queues;
1233 const unsigned int nb_rx_queues = rx_queues->num_of_queue;
deref_ptr: Directly dereferencing pointer tx_queues.
1234 const unsigned int nb_tx_queues = tx_queues->num_of_queue;
1235 unsigned int i;
1236
1237 /* do some parameter checking */
CID 345004: Dereference before null check (REVERSE_INULL)
[select issue]
1238 if (rx_queues == NULL && nb_rx_queues > 0)
1239 return -1;
CID 345029 (#1 of 1): Dereference before null check (REVERSE_INULL)
check_after_deref: Null-checking tx_queues suggests that it may be
null, but it has already been dereferenced on all paths leading to
the check.
1240 if (tx_queues == NULL && nb_tx_queues > 0)
1241 return -1;
Coverity issue: 345029
Coverity issue: 345044 Fixes: da6ba28f0540 ("net/pcap: use a struct to pass user options") Cc: stable@dpdk.org Signed-off-by: Kevin Traynor <ktraynor@redhat.com> Acked-by: Cian Ferriter <cian.ferriter@intel.com>
Thomas Monjalon [Wed, 6 Nov 2019 12:09:43 +0000 (13:09 +0100)]
cmdline: replace FreeBSD ifdef for IP address parsing
The constants like AF_INET are in sys/socket.h in FreeBSD.
The #ifdef macro __FreeBSD__ is replaced with RTE_EXEC_ENV_FREEBSD
in order to be consistent across DPDK files, and allow to grep
for EXEC_ENV among other benefits.
During LTO build compiler reports some 'false positive' warnings about
variables being possibly used uninitialized. This patch silences these
warnings.
Exemplary compiler warning to suppress (with LTO enabled):
error: ‘transceiver_type’ may be used uninitialized in this function
[-Werror=maybe-uninitialized]
switch (transceiver_type) {
Signed-off-by: Andrzej Ostruszka <aostruszka@marvell.com>
During LTO build compiler reports some 'false positive' warnings about
variables being possibly used uninitialized. This patch silences these
warnings.
Exemplary compiler warning to suppress (with LTO enabled):
error: ‘features’ may be used uninitialized in this function
[-Werror=maybe-uninitialized]
if (RTE_VHOST_NEED_LOG(features)) {
Signed-off-by: Andrzej Ostruszka <aostruszka@marvell.com>
During LTO build compiler reports some 'false positive' warnings about
variables being possibly used uninitialized. This patch silences these
warnings.
Exemplary compiler warning to suppress (with LTO enabled):
error: ‘filter_idx’ may be used uninitialized in this function
[-Werror=maybe-uninitialized]
PMD_DRV_LOG(INFO, "Added port %d with AQ command with index %d",
Signed-off-by: Andrzej Ostruszka <aostruszka@marvell.com>
During LTO build compiler reports some 'false positive' warnings about
variables being possibly used uninitialized. This patch silences these
warnings.
Exemplary compiler warning to suppress (with LTO enabled):
error: ‘link’ may be used uninitialized in this function
[-Werror=maybe-uninitialized]
if (link) {
Signed-off-by: Andrzej Ostruszka <aostruszka@marvell.com>
This patch fixes 'maybe-uninitialized' warnings reported by compiler
when using LTO.
Compiler warning pointing to this error (with LTO enabled):
error: ‘kg_cfg.extracts[0].masks[0].mask’ may be used uninitialized in
this function [-Werror=maybe-uninitialized]
extr->masks[j].mask = cfg->extracts[i].masks[j].mask;
Fixes: 16bbc98a3e63 ("bus/fslmc: update MC to 10.3.x") Cc: stable@dpdk.org Signed-off-by: Andrzej Ostruszka <aostruszka@marvell.com>
During LTO build compiler reports some 'false positive' warnings about
variables being possibly used uninitialized. This patch silences these
warnings.
Exemplary compiler warning to suppress (with LTO enabled):
error: ‘stats.greatest_free_size’ may be used uninitialized in this
function [-Werror=maybe-uninitialized]
return len - overhead;
Signed-off-by: Andrzej Ostruszka <aostruszka@marvell.com> Acked-by: Yipeng Wang <yipeng1.wang@intel.com>
During LTO build compiler reports some 'false positive' warnings about
variables being possibly used uninitialized. This patch silences these
warnings.
Exemplary compiler warning to suppress (with LTO enabled):
error: ‘chunk’ may be used uninitialized in this function
[-Werror=maybe-uninitialized]
bkt->current_chunk = (uintptr_t)chunk;
Signed-off-by: Andrzej Ostruszka <aostruszka@marvell.com>
During LTO build compiler reports some 'false positive' warnings about
variables being possibly used uninitialized. This patch silences these
warnings.
Exemplary compiler warning to suppress (with LTO enabled):
error: ‘service_id’ may be used uninitialized in this function
[-Werror=maybe-uninitialized]
ret = evt_service_setup(service_id);
Signed-off-by: Andrzej Ostruszka <aostruszka@marvell.com> Reviewed-by: Jerin Jacob <jerinj@marvell.com>
Fix the logic for the case of event queue allowing all schedule types.
Compiler warning pointing to this error (with LTO enabled):
error: ‘sched_type’ may be used uninitialized in this function
[-Werror=maybe-uninitialized]
if ((ret < 0 && ret != -EOVERFLOW) ||
Fixes: 6750b21bd6af ("eventdev: add default software timer adapter") Cc: stable@dpdk.org Signed-off-by: Andrzej Ostruszka <aostruszka@marvell.com> Reviewed-by: Jerin Jacob <jerinj@marvell.com>
This patch adds an option to enable link time optimization. In addition
to LTO option itself (-flto) fat-lto-objects are being used. This is
because during the build pmdinfogen scans the generated ELF objects to
find this_pmd_name* symbol in symbol table. Without fat-lto-objects gcc
produces ELF only with extra symbols for internal use during linking.
Signed-off-by: Andrzej Ostruszka <aostruszka@marvell.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Every implementation of a particular version of given symbol needs to be
marked in its declaration as such (using `__vsym` macro). This patch
fixes this and also clarifies the documentation about that.
Signed-off-by: Andrzej Ostruszka <aostruszka@marvell.com> Acked-by: Neil Horman <nhorman@tuxdriver.com>
test/atomic: add atomic 16/32/64-bit exchanges test
The test works by creating a token comprised of random data
and a CRC8 value, using the rte_atomicXX_exchange to exchange
the new token for a previously generated token, and then
verifying that the exchanged data is intact (i.e. the CRC8
is still correct for the data).
Signed-off-by: David Christensen <drc@linux.vnet.ibm.com> Acked-by: David Marchand <david.marchand@redhat.com>
Anatoly Burakov [Tue, 5 Nov 2019 15:15:12 +0000 (15:15 +0000)]
vfio: fix DMA mapping of external heaps
Currently, externally created heaps are supposed to be automatically
mapped for VFIO DMA by EAL, however they only do so if, at the time of
heap creation, VFIO is initialized and has at least one device
available. If no devices are available at the time of heap creation (or
if devices were available, but were since hot-unplugged, thus dropping
all VFIO container mappings), then VFIO mapping code would have skipped
over externally allocated heaps.
The fix is two-fold. First, we allow externally allocated memory
segments to be marked as "heap" segments. This allows us to distinguish
between external memory segments that were created via heap API, from
those that were created via rte_extmem_register() API.
Then, we fix the VFIO code to only skip non-heap external segments.
Also, since external heaps are not guaranteed to have valid IOVA
addresses, we will skip those which have invalid IOVA addresses as well.
Fixes: 0f526d674f8e ("malloc: separate creating memseg list and malloc heap") Cc: stable@dpdk.org Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com> Tested-by: Rajesh Ravi <rajesh.ravi@broadcom.com> Acked-by: David Marchand <david.marchand@redhat.com>
Anatoly Burakov [Mon, 4 Nov 2019 17:41:33 +0000 (17:41 +0000)]
vfio: fix DMA mapping with default container
When requesting DMA mapping to default container, we are meant to
supply the RTE_VFIO_DEFAULT_CONTAINER_FD value, however this is
not handled correctly by get_vfio_cfg_by_container_fd(), because
it only looks at actual fd values and does not check for this
special case.
Fix it to return default container if the fd requested is the
special RTE_VFIO_DEFAULT_CONTAINER_FD value.
Fixes: 4106d89a18f8 ("vfio: allow DMA map to the default container") Cc: stable@dpdk.org Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com> Reviewed-by: David Marchand <david.marchand@redhat.com> Acked-by: Shahaf Shuler <shahafs@mellanox.com>
Sunil Kumar Kori [Wed, 30 Oct 2019 16:26:44 +0000 (21:56 +0530)]
examples/l2fwd-event: add infra to split eventdev framework
Add infra to split eventdev framework based on event Tx adapter
capability.
If event Tx adapter has internal port capability then we use
`rte_event_eth_tx_adapter_enqueue` to transmitting packets else
we use a SINGLE_LINK event queue to enqueue packets to a service
core which is responsible for transmitting packets.
Signed-off-by: Sunil Kumar Kori <skori@marvell.com> Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com> Acked-by: Nipun Gupta <nipun.gupta@nxp.com> Acked-by: Jerin Jacob <jerinj@marvell.com>
Pavan Nikhilesh [Wed, 30 Oct 2019 16:26:43 +0000 (21:56 +0530)]
examples/l2fwd-event: add infra for eventdev
Add infra to select event device as a mode to process packets through
command line arguments. Also, allow the user to select the schedule type
to be RTE_SCHED_TYPE_ORDERED, RTE_SCHED_TYPE_ATOMIC or
RTE_SCHED_TYPE_PARALLEL.
Usage:
`--mode="eventdev"` or `--mode="poll"`
`--eventq-sched="ordered"`, `--eventq-sched="atomic"` or
`--event-sched=parallel`
Signed-off-by: Sunil Kumar Kori <skori@marvell.com> Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com> Acked-by: Nipun Gupta <nipun.gupta@nxp.com> Acked-by: Jerin Jacob <jerinj@marvell.com>
Hemant Agrawal [Tue, 5 Nov 2019 07:59:06 +0000 (13:29 +0530)]
config: remove redundant DPAA2 build
dpaa and dpaa2 config have evolved to be same. The same binary
can now work across the platforms. So, there is no need to maintain
two different build configs.
The dpaa config shall work for both generation of dpaa platforms.
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com> Reviewed-by: David Marchand <david.marchand@redhat.com>
Rasesh Mody [Wed, 6 Nov 2019 05:45:20 +0000 (21:45 -0800)]
net/bnx2x: fix meson build
Use kernel headers for __le* types to avoid potential conflicts
resulting in redefinition errors for some Linux build environments.
Add check for FreeBSD execution environments.
Without this fix, aarch64 builds can fail with error below:
In file included from ../drivers/net/bnx2x/bnx2x.h:22,
from ../drivers/net/bnx2x/bnx2x_ethdev.c:8:
../drivers/net/bnx2x/bnx2x_osal.h:27:17: error:
conflicting types for ‘uint64_t’
#define __le64 uint64_t
^~~~~~~~
In file included from /usr/include/stdint.h:37,
from /usr/lib/gcc/aarch64-linux-gnu/8/include/stdint.h:9,
from ../lib/librte_eal/common/include/arch/arm/rte_byteorder.h:16,
from ../drivers/net/bnx2x/bnx2x.h:17,
from ../drivers/net/bnx2x/bnx2x_ethdev.c:8:
/usr/include/aarch64-linux-gnu/bits/stdint-uintn.h:27:20: note:
previous declaration of ‘uint64_t’ was here
typedef __uint64_t uint64_t;
^~~~~~~~
There is also one minor change mixed in this commit:
Some use of __FreeBSD__ are replaced by RTE_EXEC_ENV_FREEBSD.
Olivier Matz [Mon, 4 Nov 2019 10:06:23 +0000 (11:06 +0100)]
test/mempool: fix false positive result
The ret variable, initialized to -1, is changed to 0 during the test,
making the test successful in some cases where it should return a
failure.
Fix this by always using the GOTO_ERR() macro that sets the ret
variable before doing the goto.
Fixes: 923ceaeac140 ("test/mempool: add unit test cases") Signed-off-by: Olivier Matz <olivier.matz@6wind.com> Acked-by: Andrew Rybchenko <arybchenko@solarflare.com>
Olivier Matz [Tue, 5 Nov 2019 15:37:06 +0000 (16:37 +0100)]
mempool: use specific macro for object alignment
For consistency, RTE_MEMPOOL_ALIGN should be used in place of
RTE_CACHE_LINE_SIZE. They have the same value, because the only arch
that was defining a specific value for it has been removed from DPDK.
The previous commit reduced the amount of required memory when
populating the mempool with non IOVA-contiguous memory.
Since there is no big advantage to have a fully iova-contiguous mempool
if it is not explicitly asked, remove this code, it simplifies the
populate function.
Olivier Matz [Tue, 5 Nov 2019 15:37:01 +0000 (16:37 +0100)]
mempool: reduce wasted space on populate
The size returned by rte_mempool_op_calc_mem_size_default() is aligned
to the specified page size. Therefore, with big pages, the returned size
can be much more that what we really need to populate the mempool.
For instance, populating a mempool that requires 1.1GB of memory with
1GB hugepages can result in allocating 2GB of memory.
This problem is hidden most of the time due to the allocation method of
rte_mempool_populate_default(): when try_iova_contig_mempool=true, it
first tries to allocate an iova contiguous area, without the alignment
constraint. If it fails, it fallbacks to an aligned allocation that does
not require to be iova-contiguous. This can also fallback into several
smaller aligned allocations.
This commit changes rte_mempool_op_calc_mem_size_default() to relax the
alignment constraint to a cache line and to return a smaller size.
Hemant Agrawal [Tue, 5 Nov 2019 14:23:20 +0000 (19:53 +0530)]
bus/dpaa: move QBMAN global init to bus
DPAA SEC shall be able to work independent of DPAA ETH
driver.
This patch moves qbman init to bus, so that any driver
can use them even when no eth resources are present
or none of the eth devices are probed.
Radu Bulie [Tue, 5 Nov 2019 14:23:19 +0000 (19:53 +0530)]
mempool/dpaa2: report error on endless loop in mbuf release
When BMAN is not able to accept more buffers, it could be that
there are no FBPR's (internal mem provided to bman) left.
Report error in such condition.
git.droids-corp.org / dpdk.git / log