Shougang Wang [Wed, 13 Nov 2019 01:37:56 +0000 (01:37 +0000)]
net/ice: fix flow director profile removal
The removal of FDIR profile should start from the next
of ICE_FLTR_PTYPE_NONF_NONE.
Fixes: 109e8e06249e ("net/ice: configure HW flow director rule") Signed-off-by: Shougang Wang <shougangx.wang@intel.com> Acked-by: Qiming Yang <qiming.yang@intel.com> Acked-by: Xiaolong Ye <xiaolong.ye@intel.com>
Shougang Wang [Wed, 13 Nov 2019 01:37:55 +0000 (01:37 +0000)]
net/ice: fix memzone reserve and release in flow director
To avoid memzone reserve failure and memory leak, following
resources management should be added.
- Check if the FDIR Memzone already exists before reserving.
- Free FDIR memzone when teardown and other failure scenarios.
Fixes: 84dc7a95a2d3 ("net/ice: enable flow director engine") Signed-off-by: Shougang Wang <shougangx.wang@intel.com> Acked-by: Qi Zhang <qi.z.zhang@intel.com> Acked-by: Xiaolong Ye <xiaolong.ye@intel.com>
Bruce Richardson [Tue, 12 Nov 2019 13:40:23 +0000 (13:40 +0000)]
net/i40e: fix clang build with 16B descriptors
When compiling with 16B descriptor support enabled, clang compiles gave
an error, complaining that the final parameter of _mm256_blend_epi32()
had to be an immediate value (i.e. compile-time constant):
i40e_rxtx_vec_avx2.c:561:21: error: argument to
'__builtin_ia32_pblendd256' must be a constant integer
__m256i tmp0_1 = _mm256_blend_epi32(fdir_zero_mask,
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
While it appears that GCC was able to convert the constant variable
value "fdir_blend_mask" into the blend call, clang was not doing so. To
guarantee the use of an immediate we convert the variable value to a
"#define".
Fixes: 7d087a0a8b8e ("net/i40e: support flow director on AVX Rx") Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Xiaolong Ye <xiaolong.ye@intel.com>
Chenxu Di [Tue, 12 Nov 2019 05:56:06 +0000 (05:56 +0000)]
net/i40e: fix hotplug remove
testpmd will occur infinite loops when device hotplug remove.
We can fix the issue by using the pci generic remove function
Fixes: ac89d46096d5 ("net/i40e: release port upon close") Signed-off-by: Chenxu Di <chenxux.di@intel.com> Acked-by: Xiaolong Ye <xiaolong.ye@intel.com>
Chenxu Di [Tue, 12 Nov 2019 06:00:14 +0000 (06:00 +0000)]
net/ixgbe: fix hotplug remove
testpmd will occur infinite loops when device hotplug remove.
We can fix the issue by using the pci generic remove function
Fixes: f2f4990eff94 ("net/ixgbe: release port upon close") Signed-off-by: Chenxu Di <chenxux.di@intel.com> Acked-by: Xiaolong Ye <xiaolong.ye@intel.com>
Simei Su [Thu, 7 Nov 2019 05:47:03 +0000 (13:47 +0800)]
net/ice: fix crash with wrong package
This patch fixes core dump issue when entering safe mode with a
wrong ice.pkg. In safe mode, rte_flow is not supported and it
won't initialize any flow engine.
Fixes: 7615a6895009 ("net/ice: rework for generic flow enabling") Signed-off-by: Simei Su <simei.su@intel.com> Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Qiming Yang [Mon, 11 Nov 2019 03:24:24 +0000 (11:24 +0800)]
net/ice: fix link status recovery
This patch fixes a kernel driver link status issue by recovering
link status when device stops.
Fixes: e6161345d8a9 ("net/ice: support link status change") Cc: stable@dpdk.org Signed-off-by: Qiming Yang <qiming.yang@intel.com> Acked-by: Xiaolong Ye <xiaolong.ye@intel.com>
Simei Su [Mon, 11 Nov 2019 07:55:27 +0000 (15:55 +0800)]
net/ice: fix RSS rule destroy
This patch changes RSS rule destroy interface from ice_rem_vsi_rss_cfg()
to ice_rem_rss_cfg(). To coordinate with input set change, it should
destroy a specific flow rule but not all vsi cfg.
Dekel Peled [Mon, 11 Nov 2019 14:32:31 +0000 (16:32 +0200)]
net/mlx5: fix check of RSS queue index
RSS action validation function checks the queues included in RSS
to make sure they are valid.
A Queue is considered valid if the pointer to the queue (item at
location queue-index of RxQ array) is not a null value.
The queue indices are not checked. If a large value is entered as
queue index, using it as an index in RxQ array will result in a
pointer to memory out of array bounds. If this memory contains a
value which is not null, this queue will be wrongly considered valid.
This patch updates function mlx5_flow_validate_action_rss() with
check of the input queue indices, as done in function
mlx5_flow_validate_action_queue().
Host adapter attributes are queried from kernel via multiple
DevX calls. The retrieved data were erroneously overwritten,
the order of querying is fixed.
Fixes: 6bc327b94fe8 ("net/mlx5: fill meter capabilities using DevX") Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com> Acked-by: Matan Azrad <matan@mellanox.com>
Akhil Goyal [Wed, 20 Nov 2019 05:58:45 +0000 (11:28 +0530)]
crypto/openssl: support in-place scatter gather
As per current support, Scatter Gather is only supported
for out of place input and output buffers.
This patch add support for Scatter Gather for in-place buffers.
Signed-off-by: Akhil Goyal <akhil.goyal@nxp.com> Acked-by: Anoob Joseph <anoobj@marvell.com>
Fan Zhang [Wed, 6 Nov 2019 10:54:25 +0000 (10:54 +0000)]
examples/fips_validation: fix auth verify
Fixes: f64adb6714e0 ("examples/fips_validation: support HMAC parsing") Cc: stable@dpdk.org
This patch fixes the incorrect mbuf write and digest memory leak in
fips_validation authentication verify.
Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com> Acked-by: Marko Kovacevic <marko.kovacevic@intel.com>
Pablo de Lara [Thu, 14 Nov 2019 12:21:04 +0000 (12:21 +0000)]
doc: update release notes for AESNI PMDs
AESNI MB and AESNI GCM PMDs were updated to support
latest IPSec Multi buffer library (v0.53), and the user guide
of hose drivers were updated but not release notes.
Fixes: 4b701523742e ("crypto/aesni_gcm: support in-place chained mbufs") Fixes: d91dc5835656 ("doc: support IPsec Multi-buffer lib v0.53") Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Lukasz Bartosik [Wed, 6 Nov 2019 15:48:14 +0000 (16:48 +0100)]
examples/ipsec-secgw: fix default configuration
Update default configuration of ipsec-secgw:
1.In ep0.cfg change SPI value used by two inbound IPv6 security
policies from 15 to 115 and 16 to 116 to point to existing inbound
SAs. There are no inbound SAs with SPI value 15, 16.
- In ep1.cfg change SPI value used by two outbound IPv6 security
policies from 15 to 115 and 16 to 116 to point to existing outbound
SAs. There are no outbound SAs with SPI value 15, 16. Add missing
priority parameter in two inbound IPv4 security policies.
Xueming Li [Tue, 12 Nov 2019 14:50:28 +0000 (14:50 +0000)]
malloc: fix realloc padded element size
When resize a memory with next element, the original element size grows.
If the orginal element has padding, the real inner element size didn't
grow as well and this causes trailer verification failure when malloc
debug enabled.
Fixes: af75078fece3 ("first public release") Cc: stable@dpdk.org Signed-off-by: Xueming Li <xuemingl@mellanox.com> Reviewed-by: Anatoly Burakov <anatoly.burakov@intel.com>
Xueming Li [Tue, 12 Nov 2019 14:50:27 +0000 (14:50 +0000)]
malloc: fix realloc copy size
In rte_realloc, if the old element has pad and need to allocate a new
memory, the padding size was not deducted, so more data was copied to
new data area.
Fixes: af75078fece3 ("first public release") Cc: stable@dpdk.org Signed-off-by: Xueming Li <xuemingl@mellanox.com> Reviewed-by: Anatoly Burakov <anatoly.burakov@intel.com>
Bruce Richardson [Tue, 12 Nov 2019 16:41:36 +0000 (16:41 +0000)]
kernel/freebsd: always use clang for kmod compilation
Clang is the system compiler for FreeBSD and kernel module builds can fail
when built with gcc, e.g. when testing with test-meson-builds.sh.
Therefore, it's safer to always use clang to build the kmods since the
actual flags used are outside of DPDK's control and cannot be guaranteed to
work with all compilers.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Luca Boccassi <bluca@debian.org>
Bruce Richardson [Tue, 12 Nov 2019 16:41:35 +0000 (16:41 +0000)]
kernel/freebsd: allow installing kernel modules
Set the install path for the kernel modules as /boot/modules. This may
ease the integration with the official FreeBSD ports system as all
components should be correctly located in the staging directory after
running "ninja install"
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Luca Boccassi <bluca@debian.org>
Bruce Richardson [Tue, 19 Nov 2019 15:15:43 +0000 (15:15 +0000)]
ci: add 32-bit travis builds
Add a travis job to build for 32-bit on 64-bit systems to catch additional
build errors, for example, incorrect use of printf specifiers with uint64_t
types.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Aaron Conole <aconole@redhat.com>
David Marchand [Tue, 19 Nov 2019 12:59:23 +0000 (13:59 +0100)]
build: remove unneeded meson option
The meson option has been missed when removing this code.
Fixes: 8e35792c5325 ("eal: remove dead code on NUMA node detection") Cc: stable@dpdk.org Signed-off-by: David Marchand <david.marchand@redhat.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Michael Pfeiffer [Fri, 15 Nov 2019 13:03:28 +0000 (14:03 +0100)]
kni: reduce interface name size
The name in rte_kni_device_info is passed to the kernel, which allows
interface names with at most 16 bytes (IFNAMSIZ). rte_kni_alloc with a
longer name currently trigger a kernel BUG in alloc_netdev_mqs in
net/core/dev.c. Reduce RTE_KNI_NAMESIZE to prevent this situation.
Signed-off-by: Michael Pfeiffer <michael.pfeiffer@tu-ilmenau.de> Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Anatoly Burakov [Thu, 14 Nov 2019 13:58:21 +0000 (13:58 +0000)]
mempool: remove check for bad IOVA when populating
Currently, mempool will check if IOVA is bad for a segment, and reject
the IOVA if hugepages are also enabled. This check is wrong because now
that we have external memory segments, they are allowed to have their
IOVA's to be invalid. This check also doesn't make much sense in the
first place, because the following code can handle bad IOVA's perfectly
well (and in fact, this check is not triggering a failure when
--no-huge option is enabled), so there is not much sense to check for
this in the first place.
Anatoly Burakov [Thu, 14 Nov 2019 13:58:20 +0000 (13:58 +0000)]
mempool: use actual IOVA addresses when populating
Currently, when mempool is being populated, we get IOVA address
of every segment using rte_mem_virt2iova(). This works for internal
memory, but does not really work for external memory, and does not
work on platforms which return RTE_BAD_IOVA as a result of this
call (such as FreeBSD). Moreover, even when it works, the function
in question will do unnecessary pagewalks in IOVA as PA mode, as
it falls back to rte_mem_virt2phy() instead of just doing a lookup in
internal memseg table.
To fix it, replace the call to first attempt to look through the
internal memseg table (this takes care of internal and external memory),
and fall back to rte_mem_virt2iova() when unable to perform VA->IOVA
translation via memseg table.
Vamsi Attunuru [Sun, 17 Nov 2019 15:12:44 +0000 (16:12 +0100)]
eal/linux: remove KNI restriction on IOVA
Now that KNI supports VA (with kernel versions starting 4.6.0), we can
accept IOVA as VA, but KNI must be configured for this.
Pass iova_mode when creating KNI netdevs.
So far, IOVA detection policy forced IOVA as PA when KNI is loaded,
whatever the buses IOVA requirements were.
We can now use IOVA as VA, but this comes with a cost in KNI.
When no constraint is expressed by the buses, keep the current behavior
of choosing PA.
Note: this change supposes that dpdk is built on the same kernel than
the target system kernel; no objection has been expressed on this topic.
Signed-off-by: Vamsi Attunuru <vattunuru@marvell.com> Signed-off-by: Kiran Kumar K <kirankumark@marvell.com> Signed-off-by: David Marchand <david.marchand@redhat.com> Reviewed-by: Jerin Jacob <jerinj@marvell.com>
Vamsi Attunuru [Sun, 17 Nov 2019 15:12:43 +0000 (16:12 +0100)]
kni: support userspace VA
Patch adds support for kernel module to work in IOVA = VA mode by
providing address translation routines to convert userspace VA to
kernel VA.
KNI performance using PA is not changed by this patch.
But comparing KNI using PA to KNI using VA, the latter will have lower
performance due to the cost of the added translation.
This translation is implemented only with kernel versions starting 4.6.0.
Signed-off-by: Vamsi Attunuru <vattunuru@marvell.com> Signed-off-by: Kiran Kumar K <kirankumark@marvell.com> Reviewed-by: Jerin Jacob <jerinj@marvell.com>
Zhike Wang [Wed, 13 Nov 2019 11:03:28 +0000 (19:03 +0800)]
vhost: fix vring requests validation broken if no FD
When VHOST_USER_VRING_NOFD_MASK is set, the fd_num is 0,
so validate_msg_fds() will return error. In this case,
the negotiation of vring message between vhost user front end and
back end would fail, and as a result, vhost user link could NOT be up.
How to reproduce:
1.Run dpdk testpmd insides VM, which locates at host with ovs+dpdk.
2.Notice that inside ovs there are endless logs regarding failure to
handle VHOST_USER_SET_VRING_CALL, and link of vm could NOT be up.
Fixes: bf472259dde6 ("vhost: fix possible denial of service by leaking FDs") Cc: stable@dpdk.org Signed-off-by: Zhike Wang <wangzk320@163.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> Reviewed-by: Tiwei Bie <tiwei.bie@intel.com>
Xiao Wang [Fri, 18 Oct 2019 00:01:47 +0000 (20:01 -0400)]
doc: fix address type description for IOAT rawdev
Align the description to what the code snippet shows.
Fixes: 0a92e63fc4cd ("raw/ioat: add local API to perform copies") Cc: stable@dpdk.org Signed-off-by: Xiao Wang <xiao.w.wang@intel.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
David Marchand [Tue, 8 Oct 2019 09:47:36 +0000 (11:47 +0200)]
doc: fix internal links for older releases
Using external explicit references to http://doc.dpdk.org makes older
releases documentation point to the current master documentation pages.
Switch to internal references.
Fixes: 59ad25fe2184 ("doc: add overview of qat guide") Fixes: 30e7fbd62839 ("doc: add event timer adapter guide") Fixes: b7f859c9a9a5 ("doc: add switch representation documentation") Fixes: f714a18885a6 ("app/testbbdev: add test application for bbdev") Cc: stable@dpdk.org Signed-off-by: David Marchand <david.marchand@redhat.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
The OSS-security project functions as a single point of contact for
pre-release, embargoed security notifications. Distributions and major
vendors are subscribed to this private list, so that they can be warned
in advance and schedule the work required to fix the vulnerability.
List and link this process in the DPDK security process document.
Signed-off-by: Luca Boccassi <luca.boccassi@microsoft.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Krzysztof Kanas [Tue, 12 Nov 2019 20:31:02 +0000 (21:31 +0100)]
test: optimise fd closing in forks
Caught while investigating timeouts on a ARM64 server.
Stracing a test process running the eal_flags_autotest, we can see that
the fork helper is checking all possible file descriptors from
getdtablesize() to 2, and close the existing ones.
We can do better by inspecting this forked process /proc/self/fd
directory.
Besides, checking file descriptors via /proc/self/fd only makes sense for
Linux. This code was a noop on FreeBSD.
Fixes: af75078fece3 ("first public release") Cc: stable@dpdk.org Signed-off-by: Krzysztof Kanas <kkanas@marvell.com> Signed-off-by: David Marchand <david.marchand@redhat.com> Tested-by: Krzysztof Kanas <kkanas@marvell.com> Acked-by: Kevin Traynor <ktraynor@redhat.com>
Wangyu (Eric) [Wed, 13 Nov 2019 07:17:30 +0000 (07:17 +0000)]
bus/pci: align next mapping address on page boundary
Currently, the next address picked by PCI mapping infrastructure
may be page-unaligned due to BAR length being smaller than page size.
This leads to a situation where the requested map address is invalid,
resulting in mmap() call returning an arbitrary address,
which will later interfere with device BAR mapping in secondary processes.
Fix it by always aligning the next requested address on page boundary.
Fixes: c752998b5e2e ("pci: introduce library and driver") Cc: stable@dpdk.org Signed-off-by: Xiaofeng Deng <dengxiaofeng@huawei.com> Signed-off-by: Wangyu (Eric) <seven.wangyu@huawei.com> Acked-by: Wei Hu (Xavier) <xavier.huwei@huawei.com> Acked-by: Min Hu (Connor) <humin29@huawei.com> Acked-by: Anatoly Burakov <anatoly.burakov@intel.com> Acked-by: Gavin Hu <gavin.hu@arm.com>
The #ifdef to conditionally include <sys/socket.h> on BSD
is unnecessary. It is harmless to include the header on other
OS's. An extra include is better than an #ifdef.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Olivier Matz <olivier.matz@6wind.com>
David Marchand [Tue, 12 Nov 2019 12:37:33 +0000 (13:37 +0100)]
examples/l2fwd: fix build warning with system wide install
Caught when compiling this example with pkg-config:
## Building l2fwd
...
main.c: In function ‘main’:
main.c:716:3: warning: ‘rte_eth_dev_set_ptypes’ is deprecated: Symbol
is not yet part of stable ABI [-Wdeprecated-declarations]
716 | ret = rte_eth_dev_set_ptypes(portid, RTE_PTYPE_UNKNOWN, NULL,
| ^~~
In file included from main.c:38:
...build-x86-default/install-root/usr/local/include/rte_ethdev.h:2661:5:
note: declared here
2661 | int rte_eth_dev_set_ptypes(uint16_t port_id, uint32_t
ptype_mask,
| ^~~~~~~~~~~~~~~~~~~~~~
ln -sf l2fwd-shared build/l2fwd
Fixes: 9731df2e7554 ("examples/l2fwd: disable packet type parsing") Signed-off-by: David Marchand <david.marchand@redhat.com> Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Maxime Coquelin [Tue, 3 Sep 2019 15:34:22 +0000 (17:34 +0200)]
vhost: fix possible denial of service by leaking FDs
A malicious Vhost-user master could send in loop hand-crafted
vhost-user messages containing more file descriptors the
vhost-user slave expects. Doing so causes the application using
the vhost-user library to run out of FDs.
This issue has been assigned CVE-2019-14818
Fixes: 8f972312b8f4 ("vhost: support vhost-user") Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Maxime Coquelin [Fri, 23 Aug 2019 13:17:05 +0000 (15:17 +0200)]
vhost: fix possible denial of service on SET_VRING_NUM
vhost_user_set_vring_num() performs multiple allocations
without checking whether data were previously allocated.
It may cause a denial of service because of the memory leaks
that happen if a malicious vhost-user master keeps sending
VHOST_USER_SET_VRING_NUM request until the slave runs out
of memory.
This issue has been assigned CVE-2019-14818
Fixes: b0a985d1f340 ("vhost: add dequeue zero copy") Reported-by: Jason Wang <jasowang@redhat.com> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Ray Kinsella [Mon, 11 Nov 2019 11:57:57 +0000 (11:57 +0000)]
doc: introduce major ABI versions
This policy change introduces major ABI versions, these are
declared every year, typically aligned with the LTS release
and are supported by subsequent releases in the following year.
This change is intended to improve ABI stabilty for those projects
consuming DPDK.
Signed-off-by: Ray Kinsella <mdr@ashroe.eu> Acked-by: John McNamara <john.mcnamara@intel.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas@monjalon.net>
Ray Kinsella [Mon, 11 Nov 2019 11:57:56 +0000 (11:57 +0000)]
doc: separate versioning guide into version and policy
Separate versioning.rst into abi versioning and abi policy guidance, in
preparation for adding more detail to the abi policy. Add an entry to the
maintainer file for the abi policy.
Signed-off-by: Ray Kinsella <mdr@ashroe.eu> Acked-by: John McNamara <john.mcnamara@intel.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
David Hunt [Tue, 29 Oct 2019 11:40:13 +0000 (11:40 +0000)]
examples/vm_power: fix no port in guest
If there are no ports available to the guest cli application, it will
exit when setting up the default policy because it fails to set the mac
address. This should not be the case, as this example can be used for
many other use cases that do not need ports.
If ports not found, simply set nb_mac_to_monitor in the policy to zero
and continue.
Fixes: 70febdcfd60f ("examples: check status of getting MAC address") Signed-off-by: David Hunt <david.hunt@intel.com> Acked-by: Lei Yao <lei.a.yao@intel.com>
David Hunt [Wed, 24 Jul 2019 13:18:03 +0000 (14:18 +0100)]
examples/vm_power: fix OOB frequency oscillations
The branch ratio algorithm in the vm_power_manager sample application
can be very sensitive at patricular loads in a workload, causing
oscillations between min and max frequency. For example, if a
workload is at 50%, scaling up may change the ratio
enough that it immediately thinks it needs to scale down again.
This patch introduces a sliding window recording the scale up/down
direction for the last 32 samples, and scales up if any samples indicate
we should scale up, otherwise scale down. Each core has it's own window.
Matan Azrad [Mon, 11 Nov 2019 10:42:05 +0000 (10:42 +0000)]
ethdev: fix last item detection on RSS flow expand
There is a rte_flow API which expands a RSS flow pattern to multiple
patterns according to the RSS hash types in the RSS action
configuration.
As part of the expansion, detection of the last item of the flow uses
the "next proto" field of the last configured item in the pattern list.
Wrongly, the mask of this field was not considered in order to validate
the field.
Ignore "next proto" fields when their corresponded masks invalidate them.
Fixes: fc2dd8dd492f ("ethdev: fix expand RSS flows") Cc: stable@dpdk.org Signed-off-by: Matan Azrad <matan@mellanox.com> Acked-by: Xiaoyu Min <jackmin@mellanox.com> Acked-by: Ori Kam <orika@mellanox.com>
Dekel Peled [Mon, 11 Nov 2019 17:47:35 +0000 (19:47 +0200)]
app/testpmd: set maximum LRO packet size
This patch implements use of the API for LRO aggregated packet
max size.
It adds command-line and runtime commands to configure this value,
and adds option to show the supported value.
Documentation is updated accordingly.
Dekel Peled [Mon, 11 Nov 2019 17:47:34 +0000 (19:47 +0200)]
net/mlx5: set maximum LRO packet size
This patch implements use of the API for LRO aggregated packet
max size.
Rx queue create is updated to use the relevant configuration.
Documentation is updated accordingly.
Jerin Jacob [Fri, 8 Nov 2019 16:56:03 +0000 (22:26 +0530)]
eventdev: reserve space in main structs for extension
The struct rte_eventdev and rte_eventdev_data are supposed
to be used internally only, but there is a chance that
increasing their size would break ABI for some applications.
In order to allow smooth addition of features without breaking
ABI compatibility, some space is reserved.
Sachin Saxena [Mon, 11 Nov 2019 16:08:57 +0000 (21:38 +0530)]
net/dpaa2: fix Rx offload flags on jumbo MTU set
The JUMBO frame handling in dpaa2_dev_mtu_set api was not correct.
When frame_size is greater than RTE_ETHER_MAX_LEN, the
intention is to add JUMBO flag in rx offload while it was resetting
all other flags other than JUMBO as AND operator was used instead of OR.
Fixes: 0ebce6129bc6 ("net/dpaa2: support new ethdev offload APIs") Cc: stable@dpdk.org Signed-off-by: Sachin Saxena <sachin.saxena@nxp.com> Reviewed-by: Gagandeep Singh <g.singh@nxp.com>
Thomas Monjalon [Mon, 11 Nov 2019 07:26:09 +0000 (08:26 +0100)]
ethdev: reserve space in main structs for extension
In order to allow smooth addition of features without breaking
ABI compatibility, some space is reserved in several core structs
of ethdev API.
The struct rte_eth_dev and rte_eth_dev_data are supposed
to be used internally only, but there is a chance that
increasing their size would break ABI for some applications.
Signed-off-by: Thomas Monjalon <thomas@monjalon.net> Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Raslan Darawsheh [Mon, 11 Nov 2019 11:40:20 +0000 (11:40 +0000)]
net/mlx5: set VF MAC address from host
Allow to configure the default MAC address of a VF
via its representor port in the host.
An API was proposed to specify explicitly the VF as a
target: https://patches.dpdk.org/patch/62176/
It has been rejected by the technical board in order to
keep compatibility with behavior in Intel PMDs.
http://mails.dpdk.org/archives/dev/2019-November/150588.html
Pavan Nikhilesh [Mon, 11 Nov 2019 13:19:10 +0000 (18:49 +0530)]
examples/eventdev: add new Rx RSS hash offload
Since pipeline_generic uses `rte_mbuf::hash::rss` add the new Rx offload
flag `DEV_RX_OFFLOAD_RSS_HASH` to inform PMD to copy the RSS hash result
into the mbuf.
Pavan Nikhilesh [Mon, 11 Nov 2019 13:19:07 +0000 (18:49 +0530)]
ethdev: validate offloads set by PMD
Some PMDs cannot work when certain offloads are enable/disabled, as a
workaround PMDs auto enable/disable offloads internally and expose it
through dev->data->dev_conf.rxmode.offloads.
After device specific dev_configure is called compare the requested
offloads to the offloads exposed by the PMD and, if the PMD failed
to enable a given offload then log it and return -EINVAL from
rte_eth_dev_configure, else if the PMD failed to disable a given offload
log and continue with rte_eth_dev_configure.
Suggested-by: Andrew Rybchenko <arybchenko@solarflare.com> Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com> Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>
Pavan Nikhilesh [Mon, 11 Nov 2019 13:19:06 +0000 (18:49 +0530)]
ethdev: add mbuf RSS update as an offload
Add new Rx offload flag `DEV_RX_OFFLOAD_RSS_HASH` which can be used to
enable/disable PMDs write to `rte_mbuf::hash::rss`.
PMDs notify the validity of `rte_mbuf::hash:rss` to the application
by enabling `PKT_RX_RSS_HASH ` flag in `rte_mbuf::ol_flags`.
Also update testpmd rx_offload command to include RSS_HASH
Pavan Nikhilesh [Mon, 11 Nov 2019 13:19:05 +0000 (18:49 +0530)]
ethdev: add packet type range function
Add `rte_eth_dev_set_ptypes` function that will allow the application
to inform the PMD about reduced range of packet types to handle.
Based on the ptypes set PMDs can optimize their Rx path.
-If application doesn’t want any ptype information it can call
`rte_eth_dev_set_ptypes(ethdev_id, RTE_PTYPE_UNKNOWN, NULL, 0)`
and PMD may skip packet type processing and set rte_mbuf::packet_type to
RTE_PTYPE_UNKNOWN.
-If application doesn’t call `rte_eth_dev_set_ptypes` PMD can return
`rte_mbuf::packet_type` with `rte_eth_dev_get_supported_ptypes`.
-If application is interested only in L2/L3 layer, it can inform the PMD
to update `rte_mbuf::packet_type` with L2/L3 ptype by calling
`rte_eth_dev_set_ptypes(ethdev_id,
RTE_PTYPE_L2_MASK | RTE_PTYPE_L3_MASK, NULL, 0)`.
Suggested-by: Konstantin Ananyev <konstantin.ananyev@intel.com> Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com> Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>
Haiyue Wang [Fri, 8 Nov 2019 15:44:34 +0000 (23:44 +0800)]
net/ice: optimize protocol extraction by dynamic mbuf
The original design is to use rte_mbuf::udata64 to save the metadata of
protocol extraction which has network protocol data fields and type, a
private API is used to decode this metadata.
Use the dynamic mbuf field and flags to register the needed fields in
mbuf, to avoid overwriting 'rte_mbuf::udata64', since the application
may use it. Now the protocol extraction metadate is saved into dynamic
mbuf field with 4B size, and its type and validity is indicated by the
related dynamic mbuf flags in 'rte_mbuf::ol_flags'.
Signed-off-by: Haiyue Wang <haiyue.wang@intel.com> Reviewed-by: Xiaolong Ye <xiaolong.ye@intel.com>
Bing Zhao [Fri, 8 Nov 2019 05:26:57 +0000 (07:26 +0200)]
net/mlx5: optimize tag traversal with hash list
Tag action for flow mark/flag could be reused by different flows.
When creating a new flow with mark, the existing tag resources will
be traversed in order to confirm if the action is already created.
If only one linked list is used, the searching rate will drop
significantly with the number of tag actions increasing.
By using a hash lists table, it will speed up the searching process
and in the meanwhile, the memory consumption won't be large if only
a small number tag action resources are created(compared to other
hash table implementations). The list heads array size could be
optimized with some extendable hash table in the future.
Bing Zhao [Fri, 8 Nov 2019 15:23:10 +0000 (17:23 +0200)]
net/mlx5: reorganize flow matcher resources
Matchers are created on the specific table. If a single linked list
is used to store these, then the finding process might be the
bottleneck when there are a lot of different flow matchers on a
huge amount of tables. The matchers could be move into the table
data resource structure in order to reduce the comparison times
when finding.
Bing Zhao [Fri, 8 Nov 2019 15:23:09 +0000 (17:23 +0200)]
net/mlx5: reorganize jump table resources
Jump object is associated with table object, so there is no need to
use a single linked list to store it. All the jump objects could be
put together with related flow tables.
Bing Zhao [Fri, 8 Nov 2019 15:23:08 +0000 (17:23 +0200)]
net/mlx5: reorganize flow tables with hash list
In the current flow tables organization, arrays are used. This is
fast for searching, creating related object that will be used in
flow creation. But it introduces some limitation to the table index.
Then we can reorganize the flow tables information with hash list.
When using hash list, there is no need to maintain three arrays for
NIC TX, RX and FDB tables object information.
This attribute could be used together with the table ID to generate
a 64-bits key that is unique for the hash list insertion, lookup and
deletion.
net/mlx5: control transmit doorbell register mapping
The rdma core library can map doorbell register in two ways,
depending on the environment variable "MLX5_SHUT_UP_BF":
- as regular cached memory, the variable is either missing or
set to zero. This type of mapping may cause the significant
doorbell register writing latency and requires explicit
memory write barrier to mitigate this issue and prevent
write combining.
- as non-cached memory, the variable is present and set to
not "0" value. This type of mapping may cause performance
impact under heavy loading conditions but the explicit write
memory barrier is not required and it may improve core
performance.
The new devarg is introduced "tx_db_nc", if this parameter is
set to zero, the doorbell register is forced to be mapped to
cached memory and requires explicit memory barrier after
writing to. If "tx_db_nc" is set to non-zero value the doorbell
will be mapped as non-cached memory, not requiring the memory
barrier. If "tx_db_nc" is missing the behaviour will be defined
by presence of "MLX5_SHUT_UP_BF" in environment. If variable
is missed the default value zero will be set for ARM64 hosts
and one for others.
In run time the code checks the mapping type and provides the
memory barrier after writing to tx doorbell register if it is
needed. The mapping type is extracted directly from the
uar_mmap_offset field in the queue properties.
All the other testpmd commands block access to devices that
are owned. Looks like xstat got overlooked.
Fixes: bfd5051b43b5 ("app/testpmd: new command to get extended statistics") Cc: stable@dpdk.org Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Bernard Iremonger <bernard.iremonger@intel.com>
A Vhost-blk example that support inflight feature. It uses the
new APIs that introduced in the first patch, so it can show how these
APIs work to support inflight feature.
Signed-off-by: Jin Yu <jin.yu@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Qi Zhang [Fri, 8 Nov 2019 09:40:39 +0000 (17:40 +0800)]
net/ice: free HW tables when device closed
Free the HW tables during dev_close.
Otherwise there will be two issues:
1. Memory won't be freed if the device is detached.
2. Driver can't be initialized correctly after device reset.
Fixes: a4c8c48fe3f4 ("net/ice: load OS default package") Cc: stable@dpdk.org Signed-off-by: Qi Zhang <qi.z.zhang@intel.com> Acked-by: Xiaolong Ye <xiaolong.ye@intel.com>
net/mlx5: fix asserts for multi-segment Tx offload misconfig
To support multi-segment packet sending the Tx datapath
DEV_TX_OFFLOAD_MULTI_SEGS offload must be configured.
In debug configuration the asserts arise if the multi-segment
packet is being sent and no DEV_TX_OFFLOAD_MULTI_SEGS is set.
It was found that in some execution paths asserts were missed,
this patch adds ones.
Suanming Mou [Fri, 8 Nov 2019 03:49:24 +0000 (05:49 +0200)]
net/mlx5: share tag between meter and metadata
In the meter flow split, metadata flow will be as the sub flow of meter
suffix flow. In meter suffix flow, there is already a unique id tag
exist as for the meter prefix and suffix flow match.
Make metadata feature and meter both share the unique id tag for match.
Suanming Mou [Fri, 8 Nov 2019 03:49:23 +0000 (05:49 +0200)]
net/mlx5: split meter flow
The flow with meter action will be divided to three sub flows, the
prefix flow, the meter flow and the suffix flow.
For these three sub flows, as the prefix flow and meter flow have the
meter color match. In order the make the packet from prefix flow to
match with the suffix flow, an extra metadata register is allocated.
The prefix flow will add a unique id to the register and the suffix
flow matches on that unique id.
As RSS will also divided the flow to several sub flows, flow with meter
will be divided as the sub flow of the RSS sub flows if have as below:
Original flow ->
RSS sub flow 1 ->
Meter sub flow 1 (Contain three sub flows.)
RSS sub flow 2 ->
Meter sub flow 2 (Contain three sub flows.)
......
RSS sub flow n ->
Meter sub flow n (Contain three sub flows.)
The metadata feature flow will be split as the sub flow of the meter
suffix flow.
Suanming Mou [Fri, 8 Nov 2019 03:49:21 +0000 (05:49 +0200)]
net/mlx5: add meter attach and detach
Add the meter attach and detach for the flow create.
When create the flow with meter, first try to find any created meter
action matches the flow meter id. If the meter action is already
created, just attach to it and increase the ref_cnt. If not, create
one.
For the dettach, decrease the ref_cnt, destroy the meter action while
the ref_cnt decreased to zero.
Suanming Mou [Fri, 8 Nov 2019 03:49:13 +0000 (05:49 +0200)]
net/mlx5: add policer rules operations
This commit create the color rules on the meter table for the packets.
As the prefix flow with meter action colors the packets, the packets
are transferred to the meter table with meter color match flows. Here
we create the flow rules with green yellow red actions on the meter
table. Packets match the color will be processed by the related color
flow rule.
Suanming Mou [Fri, 8 Nov 2019 03:49:12 +0000 (05:49 +0200)]
net/mlx5: prepare meter flow tables
This commit prepare the meter table and suffix table.
A flow with meter will be split to three flows. The three flows are
created on differnet tables. The packets transfer between the flows
on the tables as below:
Prefix flow -> Meter flow -> Suffix flow
Prefix flow does the user defined match and the meter action. The meter
action colors the packet and set its destination to meter table to be
processed by the meter flow.
The meter flow judges if the packet can be passed or not. If packet can
be passed, it will be transferred to the suffix table.
The suffix flow on the suffix table will apply the left user defined
actions to the packet.
The ingress egress and transfer all have the independent meter and
suffix tables.
Suanming Mou [Fri, 8 Nov 2019 03:49:09 +0000 (05:49 +0200)]
net/mlx5: allocate flow meter registers
Meter need the metadata REG_C to have the color match between the prefix
flow and the meter flow.
As the user define or metadata feature will both use the REG_C in the
suffix flow, the color match register meter uses will not impact the
register use in the later sub flow.
Another case is that tag is add before meter flow. In this case, meter
should not touch the register the tag action is using. To avoid that
case, meter should reserve the REG_C's used by user defined MLX5_APP_TAG.
git.droids-corp.org / dpdk.git / log