Harry van Haaren [Mon, 21 Aug 2017 12:58:03 +0000 (13:58 +0100)]
service: use id in lcore to service map functions
This commit updates the APIs exposed to map service cores and
services. The previous APIs required a pointer to a service,
and used two separate functions for enable and disable. The
new API uses an integer ID for the service and has a parameter
for map or unmap. Unit tests are updated and passing, and the
map file is updated to the new function names.
Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com> Acked-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
Harry van Haaren [Mon, 21 Aug 2017 12:58:02 +0000 (13:58 +0100)]
service: use id in probe and get name
This commit adds a macro to easily validate a service ID, and then
lookup the service pointer, or return a user-specified error code.
This macro will be heavily used in the following patches as it will
be ID based instead of pointer-based.
The probe_capability function is reworked to use an integer ID instead
of a pointer. Rework the service_get_name() function is updated to use
IDs. Unit tests are updated to keep things compiling after each commit.
Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com> Acked-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
Pavan Nikhilesh [Tue, 29 Aug 2017 12:32:51 +0000 (18:02 +0530)]
service: add count per lcore
This new API returns the number of services that are running on a specific
service core. It allows an application to decide which service core to run
a new service on.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com> Acked-by: Harry van Haaren <harry.van.haaren@intel.com>
This commit zeros out the elf_info struct at startup of the
pmdinfogen code. If it is not zeroed, later in the code gcc
produces "may be unused" prints. Clang does not report any
issue.
This issue is only observed when compiling pmdinfogen as an
optimized build, hence this warning is not disabled in the
existing Makefile.
This commit enables a simplification in the meson build
system, removing the requirement for "-Wno-maybe-uninitialized".
Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Wei Dai [Sun, 23 Jul 2017 09:15:12 +0000 (17:15 +0800)]
app/testpmd: add port reset command
When PF is reset, a message will show it and all its
VF need to be reset.
User can run the command "port reset port_id"
to reset the VF port and to keep same port id without
any configuration. Then user can run "port stop port_id"
and "port start port_id" to reconfigure its forwarding
mode and parameters as previous ones.
To avoid crash, current forwarding should be stopped
before running "port reset port_id".
Wei Dai [Sun, 23 Jul 2017 09:15:11 +0000 (17:15 +0800)]
net/i40e: support NIC reset
Reset a NIC by calling dev_uninit() and then dev_init().
Go through the same way in NIC PCI remove without release
of ethdev resource and then NIC PCI probe function without
ethdev resource allocation.
Wei Dai [Sun, 23 Jul 2017 09:15:10 +0000 (17:15 +0800)]
net/ixgbe: support NIC reset
Reset a NIC by calling dev_uninit and then dev_init.
Go through same way in NIC PCI remove without release of
ethdev resource and then NIC PCI probe function without
ethdev resource allocation.
Wei Dai [Sun, 23 Jul 2017 09:15:09 +0000 (17:15 +0800)]
ethdev: add NIC reset operation
This patch adds a new eth_dev layer API function rte_eth_dev_reset(),
which a DPDK application can call to reset a NIC and keep its port id
afterwards. It means that all software resources allocated in the ethdev
layer are kept, and software & hardware resources of the NIC within the
NIC's PMD are reset to a state simular to that obtained by calling the
PCI dev_uninit() and then dev_init(). This effective sequence of
dev_uninit() and dev_init() is packed into a single API function
rte_eth_dev_reset().
Please see the comments before the declaration of rte_eht_dev_reset()
in lib/librte_ether/rte_ethdev.h to get more details on why this
function is needed, what it does, when it should be called
and what an application should do after calling this function.
See also detailed explanations in the programmer's guide.
Signed-off-by: Wei Dai <wei.dai@intel.com> Reviewed-by: Remy Horton <remy.horton@intel.com>
Bruce Richardson [Thu, 31 Aug 2017 13:42:43 +0000 (14:42 +0100)]
ethdev: rename map file to match library name
While almost all libraries have the library name match the directory name,
some libraries and drivers do not. For those that don't, some have the
version filename use the libname, e.g. null_crypto, others have it use the
directory name, e.g. ether. Using the library name seems the better
standard to follow, so rename rte_ether_version.map to
rte_ethdev_version.map.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Bruce Richardson [Thu, 31 Aug 2017 13:42:42 +0000 (14:42 +0100)]
net/ring: rename map file to standardized name
The filenames of the linker map files for DPDK PMDs, all follow a
standard format: rte_pmd_<libname>_version.map. The ring version, however,
had eth instead of pmd in the name, so was non-standard. By changing
this, we no longer need the build system to explicitly be given the name of
the mapfile, as it can determine it from the directory name.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Bruce Richardson [Thu, 31 Aug 2017 13:42:41 +0000 (14:42 +0100)]
ip_frag: rename map file to standardized name
The filenames of the linker map files for DPDK libraries, all follow a
standard format: rte_<libname>_version.map. The ip_frag version, however,
was missing an underscore in the name, so was non-standard. By changing
this, we no longer need the build system to explicitly be given the name of
the mapfile, as it can determine it from the directory/library name.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Jerin Jacob [Tue, 8 Aug 2017 14:48:27 +0000 (20:18 +0530)]
eventdev: bump library version
Bumping the library version to reflect the ABI change, where
rte_event_pmd_pci_probe(), rte_event_pmd_pci_remove(),
rte_event_pmd_vdev_init(), rte_event_pmd_vdev_uninit()
functions removed from the library.
Fixes: b1b3d9f90502 ("eventdev: make vdev init and uninit functions optional") Fixes: 9a8269d56942 ("eventdev: make PCI probe and remove functions optional") Reported-by: Thomas Monjalon <thomas@monjalon.net> Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Pablo de Lara [Thu, 3 Aug 2017 01:02:36 +0000 (02:02 +0100)]
doc: announce API change in crypto driver allocation
rte_cryptodev_allocate_driver() function gets one parameter
(rte_driver), as the cryptodev_driver structure is
allocated inside the function with rte_malloc.
This function is called from a constructor function,
when crypto PMDs are registered.
If malloc fails, there is no way to recover from it,
so it is better to allocate this structure
statically, in each PMD.
Therefore, it is required to add an extra parameter in
this function, to also get a pointer to this structure.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: John McNamara <john.mcnamara@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Acked-by: Deepak Kumar Jain <deepak.k.jain@intel.com>
Pablo de Lara [Thu, 3 Aug 2017 02:12:21 +0000 (03:12 +0100)]
doc: announce crypto vdev init removal
In order to remove all dependencies on vdev for cryptodev,
the implementation of rte_cryptodev_vdev_pmd_init() function
needs to be moved to rte_cryptodev_vdev.h, and all crypto
vdevs will include it, and therefore, this function will
be removed as a public API.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: John McNamara <john.mcnamara@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Acked-by: Deepak Kumar Jain <deepak.k.jain@intel.com>
Akhil Goyal [Thu, 3 Aug 2017 15:32:11 +0000 (21:02 +0530)]
doc: announce ABI change for cryptodev/ethdev offloads
Support for security operations is planned to be added
in ethdev and cryptodev for the 17.11 release.
For this following changes are required.
- rte_cryptodev and rte_eth_dev structures need to be added
new parameter rte_security_ops which extend support for
security ops to the corresponding driver.
- rte_cryptodev_info and rte_ethd_dev_info need to be added
with rte_security_capabilities to identify the capabilities of
the corresponding driver.
Signed-off-by: Akhil Goyal <akhil.goyal@nxp.com> Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com> Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Boris Pismenny <borisp@mellanox.com> Acked-by: Shahaf Shuler <shahafs@mellanox.com>
Zhiyong Yang [Mon, 7 Aug 2017 12:42:30 +0000 (20:42 +0800)]
doc: announce ethdev port id range increase
This is an API/ABI change notice for DPDK 17.11 announcing the redefinition
of port_id. port_id is currently defined as uint8_t, which is limited to
the range 0 to 255. A larger range is required for vdev scalability.
It is necessary for a redefinition of port_id to extend it from 1 bytes to
2 bytes. All ethdev APIs and usages related to port_id will be changed at
the same time.
Signed-off-by: Zhiyong Yang <zhiyong.yang@intel.com> Acked-by: Jianfeng Tan <jianfeng.tan@intel.com> Acked-by: Yuanhan Liu <yliu@fridaylinux.org> Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com> Acked-by: Remy Horton <remy.horton@intel.com> Acked-by: Shahaf Shuler <shahafs@mellanox.com>
This flag is not needed anymore following the hotplug work done for
v17.08. It can be removed, its function is now implicitly made available
by the relevant EAL and rte_bus implementations.
Signed-off-by: Gaetan Rivet <gaetan.rivet@6wind.com> Acked-by: John McNamara <john.mcnamara@intel.com> Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com> Acked-by: Thomas Monjalon <thomas@monjalon.net>
An API/ABI change is planned for 17.11 to change following
* Remove unused flag param from rte_mempool_generic_get and _put.
* Change data type for mempool 'flag' from int to unsigned int.
Refer [1].
* Add struct rte_mempool * param into func rte_mempool_xmem_size,
rte_mempool_xmem_usage to make it mempool aware.
Refer [2].
Jerin Jacob [Tue, 11 Jul 2017 10:01:41 +0000 (15:31 +0530)]
doc: announce change to make DPDK IOVA aware
When we run DPDK on guest or VFIO mode on host,
the dpdk library or device will not be directly accessing
the physical address. Instead, the device does go through
an IO address translation memory management unit. On x86,
we call it as IOMMU and on ARM as SMMU.
More details:
http://osidays.com/osidays/wp-content/uploads/2014/12/Final_OSI2014_IOMMU_DetailedView_Sanil_Anurup.pdf
Based on discussion in the following thread
http://dpdk.org/ml/archives/dev/2017-July/070850.html
We would like to change reference to physical address to more
appropriate name as with IOMMU/SMMU with
the device won't be dealing directly with the physical address.
An ABI change is planned for 17.11 to change following
data structure or functions to more appropriate name.
Currently planned to change it iova as instead of phys
Please note: The change will be only for the name and
functional aspects of the API will remain same.
Following functions/data structures name may change.
This list is based on v17.05-rc1. It may change based on v17.11 code base.
Jianfeng Tan [Thu, 3 Aug 2017 20:38:10 +0000 (20:38 +0000)]
doc: deprecate Xen support
Following the calls on the mailing list:
http://dpdk.org/ml/archives/dev/2017-June/068151.html
The Technical Board decided to drop Xen dom0 support from EAL:
http://dpdk.org/ml/archives/dev/2017-June/068615.html
Signed-off-by: Thomas Monjalon <thomas@monjalon.net> Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com> Acked-by: John McNamara <john.mcnamara@intel.com> Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com> Acked-by: Shahaf Shuler <shahafs@mellanox.com>
Pablo de Lara [Mon, 7 Aug 2017 03:16:54 +0000 (04:16 +0100)]
app/crypto-perf: fix operation free
In crypto latency test, when some crypto operations cannot be
enqueued, they are returned to its mempool.
However, instead of freeing the operations in the array passed
to the enqueue function, the app was freeing the operations
in the array passed to the dequeue function, leading to incorrect
results.
Fixes: 15b55dd75120 ("app/crypto-perf: return crypto ops to mempool in bulk") Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Hugepages are just that, hugepages. The TLB is a sort of cache for doing
address translation and does not need to be referenced in the title, since
"TLB hugepages" doesn't make any sense (as there are no non-TLB hugepages
that you can allocate).
Fixes: 6e718ae062d0 ("doc: release notes 1.7") Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: John McNamara <john.mcnamara@intel.com>
Since modern kernels don't allow unprivileged processes to read the
pagemaps file, the instructions on running as non-root are out-of-date. Add
a note clarifying that they will only work with earlier kernel versions.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: John McNamara <john.mcnamara@intel.com>
doc: add tags and cscope make targets to quick help
Update the "make help" output to reflect the additions
of make targets for tags database generation.
Fixes: aafaea3d3b70 ("devtools: add tags and cscope index generation") Signed-off-by: Gabriel Carrillo <erik.g.carrillo@intel.com> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
David Hunt [Fri, 4 Aug 2017 10:39:02 +0000 (11:39 +0100)]
mk: add sensible default target with defconfig
Users can now use 'make defconfig' to generate a configuration using
the most appropriate defaults for the current machine.
<arch-machine-execenv-toolchain>
arch taken from uname -m
machine defaults to native
execenv is taken from uname, Linux=linuxapp, otherwise bsdapp
toolchain is taken from $CC -v to see which compiler to use
Signed-off-by: David Hunt <david.hunt@intel.com> Acked-by: Shreyansh Jain <shreyansh.jain@nxp.com> Tested-by: Jerin Jacob <jerin.jacob@caviumnetworks.com> Acked-by: Thomas Monjalon <thomas@monjalon.net>
During the refactoring of the sample app to be more generic, the
option to set -n0 and process a huge number of packets was lost.
This commit re-adds -n0, and will process INT64_MAX number of packets.
Fixes: adb5d5486c39 ("examples/eventdev_pipeline_sw_pmd: add sample app") Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
Xiaoyun Li [Fri, 4 Aug 2017 15:50:06 +0000 (23:50 +0800)]
net/i40e: revert fix of PF notify when VF not up
This reverts commit 035a8cf88f010c82de3040c895c801ce0849b065.
Don't send messages to inactive VF will cause DPDK PF failing
to send messages to kernel VF.
With this revert, this issue will be solved.
Fixes: 035a8cf88f01 ("net/i40e: fix PF notify when VF is not up") Cc: stable@dpdk.org Signed-off-by: Xiaoyun Li <xiaoyun.li@intel.com> Acked-by: Beilei Xing <beilei.xing@intel.com>
Qi Zhang [Wed, 2 Aug 2017 21:38:09 +0000 (05:38 +0800)]
net/ixgbe: fix mask flag on flow rule creation
After fail to create a flow and if this is the first flow, the
mask_added flag should be reset, or it will prevent a new flow
which require different mask be created, since the mask config
remains impact.
Thomas Monjalon [Fri, 4 Aug 2017 10:00:53 +0000 (12:00 +0200)]
eventdev: fix missing includes
The PCI helper file depends on some EAL definitions.
Spotted with devtools/check-includes.sh:
error: implicit declaration of function ‘rte_eal_process_type’
error: ‘RTE_PROC_PRIMARY’ undeclared
error: implicit declaration of function ‘rte_socket_id’
There was also this error because the inline keyword was missing:
error: ‘rte_event_pmd_pci_probe’ defined but not used
Fixes: 9a8269d56942 ("eventdev: make PCI probe and remove functions optional") Signed-off-by: Thomas Monjalon <thomas@monjalon.net> Acked-by: John McNamara <john.mcnamara@intel.com>
rte_crypto_sym.h:586:3: error: anonymous structs are a C11 extension
[-Werror,-Wc11-extensions]
rte_crypto_sym.h:510:2: error: anonymous unions are a C11 extension
[-Werror,-Wc11-extensions]
Fixes: b59502a5e3d0 ("cryptodev: add AEAD parameters in crypto operation") Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com> Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
rte_tm.h:473:2: error: anonymous unions are a C11 extension
[-Werror,-Wc11-extensions]
rte_tm.h:696:2: error: anonymous unions are a C11 extension
[-Werror,-Wc11-extensions]
rte_tm.h:904:2: error: anonymous unions are a C11 extension
[-Werror,-Wc11-extensions]
Jianfeng Tan [Wed, 2 Aug 2017 17:34:11 +0000 (17:34 +0000)]
doc: remove unnecessary deprecation notice
The original purpose of this deprecation is to make sure PCI devices
are reset whenever DPDK apps crash.
Since the commit b58eedfc7dd5 from Shijith can fix this problem without
deprecating anything, now there is no need to deprecate iomem and ioport
mapping in igb_uio.
Fixes: 3bac1dbc1ed5 ("doc: announce iomem and ioport removal from igb_uio") Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com> Acked-by: John McNamara <john.mcnamara@intel.com>
Thomas Monjalon [Sun, 16 Jul 2017 17:47:07 +0000 (19:47 +0200)]
maintainers: add event ring test to eventdev
This file was not referenced in MAINTAINERS list.
The miss is spotted with devtools/check-maintainers.sh.
As this test file is related to eventdev, they should both
have the same maintainer.
Fixes: 1ee55d7a6e4d ("test/eventdev: add auto-tests for event ring functions") Signed-off-by: Thomas Monjalon <thomas@monjalon.net> Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Thomas Monjalon [Thu, 3 Aug 2017 14:50:55 +0000 (16:50 +0200)]
doc: remove outdated list of supported OS
It is very difficult to list OS which are really supported.
Instead of continuing this unrealistic effort, a more reliable list
of tested platforms and OS is updated in the release notes.
Signed-off-by: Thomas Monjalon <thomas@monjalon.net> Acked-by: John McNamara <john.mcnamara@intel.com>
John McNamara [Thu, 3 Aug 2017 17:19:45 +0000 (18:19 +0100)]
doc: warn about UIO not working with UEFI secure boot
Added notes to the documentation warning that if UEFI secure boot
is enabled the Linux kernel may disallow the use of UIO on the
system, and a suggested workaround of using the vfio-pci kernel
module instead of igb_uio or uio_pci_generic.
Signed-off-by: John McNamara <john.mcnamara@intel.com>
Pablo de Lara [Thu, 3 Aug 2017 10:34:29 +0000 (11:34 +0100)]
test/crypto: fix error message
Fixes: 7a364faef185 ("cryptodev: remove crypto device type enumeration") Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: John McNamara <john.mcnamara@intel.com>
Pablo de Lara [Tue, 1 Aug 2017 00:33:36 +0000 (01:33 +0100)]
app/crypto-perf: fix IV allocation for AEAD
Memory is reserved after each crypto operation
for the necessary IV(s), which could be for cipher,
authentication or AEAD algorithms.
However, for AEAD algorithms (such as AES-GCM), this
memory was not being reserved, leading to potential
memory overflow.
Fixes: 8a5b494a7f99 ("app/test-crypto-perf: add AEAD parameters") Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
David Hunt [Tue, 25 Jul 2017 10:27:52 +0000 (11:27 +0100)]
app/testpmd: fix Tx first with LSC interrupt
The lsc_interrupt flag is enabled by default, and when the --tx-first is
used as a command line parameter, the 32 packets can be sent out before
the link state is up, resulting in the loss of the packets, and no further
forwarding will take place.
E.g. ./build/app/testpmd -c f0 -- --tx-first --stats-period 1
When the --tx-first is used, the lsc_interrupt flag needs to be disabled,
ensuring the links are up before forwarding traffic. Therefore, during the
parameter checking at startup, if --tx-first is used, we now warn the user,
and set lsc_interrupt to 0.
Fixes: 99cabef08855 ("app/testpmd: add parameter to start forwarding Tx first") Signed-off-by: David Hunt <david.hunt@intel.com> Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Jiayu Hu [Thu, 3 Aug 2017 09:45:19 +0000 (17:45 +0800)]
app/testpmd: fix GRO command description
In testpmd, GRO is supported by csum forwarding engine, but the cmdline
message shows GRO is supported by io forwarding engine. This patch is
to fix this issue.
net/mlx5: add parameters to enable/disable vector datapath
Vector code is very young and can present some issues for users, to avoid
them to modify the selections function by commenting the code and recompile
the PMD, new devices parameters are added to deactivate the Tx and/or Rx
vector code.
By using such device parameters, the user will be able to fall back to
regular burst functions.
Yongseok Koh [Wed, 2 Aug 2017 17:25:54 +0000 (10:25 -0700)]
net/mlx5: fix error statistics in SSE Rx
If there's a Rx completion with error (e.g, MTU mismatch), it is handled
later out of main burst loop as a slow path for performance reason.
Statistics should be corrected by subtracting counters of errored packets.
Also, the last entry of mlx5_ptype_table[] must be RTE_PTYPE_ALL_MASK to
mark error in completion.
Yongseok Koh [Tue, 1 Aug 2017 16:35:20 +0000 (09:35 -0700)]
net/mlx5: fix buffer address posting in SSE Rx
The data_off field of newly allocated mbufs is stale data. This shouldn't
be used in calculating Rx address for device when posting free buffers.
RTE_PKTMBUF_HEADROOM should be used instead and data_off of a mbuf will be
reset on packet reception anyway.
Yongseok Koh [Thu, 3 Aug 2017 21:12:40 +0000 (14:12 -0700)]
net/mlx5: fix trimming SW ring for vectorized Rx
Unlike mlx5_rx_burst(), mlx5_rx_burst_vec() doesn't replace completed
buffers one by one right after completion is processed but replenishes
multiple buffers later with rte_mempool_get_bulk(). Therefore, there could
be some buffer addresses left in the SW ring (rxq->elts[]) which have
already been delivered to application. As PMD doesn't own such buffers, it
must not be freed by PMD. "Trimming" is needed before cleanup.
A problem can be seen when quitting testpmd when
CONFIG_RTE_LIBRTE_MBUF_DEBUG=y and CONFIG_RTE_LIBRTE_MEMPOOL_DEBUG=y
Trimming should be as simple as possible, it shouldn't touch any indexes
and buffer allocation isn't necessary.
Matan Azrad [Wed, 2 Aug 2017 19:00:50 +0000 (22:00 +0300)]
net/mlx4: workaround verbs error after plug-out
Current mlx4 OFED version has bug which returns error to
ibv destroy functions when the device was plugged out, in
spite of the resources were destroyed correctly.
Hence, failsafe PMD was aborted, only in debug mode, when
it tries to remove the device in plug-out process.
The workaround added option to replace all claim_zero
assertions with debugging messages, by the way, this option
affects non ibv destroy assertions.
DPDK 18.02 release should work with Mellanox OFED-4.2 which will
include the verbs fix to this bug, then, this patch can
be removed.
Aaron Conole [Wed, 2 Aug 2017 18:02:13 +0000 (14:02 -0400)]
net/enic: fix crash when freeing 0 packet to mempool
Occasionally, the amount of packets to free from the work queue ends
perfectly on a boundary to have nb_free = 0 and pool = 0. This causes
a segfault as follows:
(gdb) bt
#0 rte_mempool_default_cache
#1 rte_mempool_put_bulk (n=0, obj_table=0x7f10deff2530, mp=0x0)
#2 enic_free_wq_bufs (wq=wq@entry=0x7efabffcd5b0,
completed_index=completed_index@entry=33)
#3 0x00007f11e9c86e17 in enic_cleanup_wq (enic=<optimized out>,
wq=wq@entry=0x7efabffcd5b0)
at /usr/src/debug/openvswitch-2.6.1/dpdk-16.11/drivers/net/enic/enic_rxtx.c:442
#4 0x00007f11e9c86e5f in enic_xmit_pkts (tx_queue=0x7efabffcd5b0,
tx_pkts=0x7f10deffb1a8, nb_pkts=<optimized out>)
at /usr/src/debug/openvswitch-2.6.1/dpdk-16.11/drivers/net/enic/enic_rxtx.c:470
#5 0x00007f11e9e147ad in rte_eth_tx_burst (nb_pkts=<optimized out>,
tx_pkts=0x7f10deffb1a8, queue_id=0, port_id=<optimized out>)
This commit makes the enic wq driver match other drivers who call the
bulk free, by checking that there are actual packets to free.
Fixes: 36935afbc53c ("net/enic: refactor Tx mbuf recycling") CC: stable@dpdk.org Reported-by: Vincent S. Cojot <vcojot@redhat.com>
Reported-at: https://bugzilla.redhat.com/show_bug.cgi?id=1468631 Signed-off-by: Aaron Conole <aconole@redhat.com> Reviewed-by: John Daley <johndale@cisco.com>
Qi Zhang [Wed, 2 Aug 2017 07:22:09 +0000 (15:22 +0800)]
net/i40e: fix Rx data segment buffer length
Buffer length be configured for each data segment should not exceed
the requested value, or device may fill data that exceed the boundary
of memory that be reserved.
Fixes: 4861cde46116 ("i40e: new poll mode driver") Cc: stable@dpdk.org Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com> Signed-off-by: Qi Zhang <qi.z.zhang@intel.com> Reviewed-by: Mark Kavanagh <mark.b.kavanagh@intel.com> Tested-by: Mark Kavanagh <mark.b.kavanagh@intel.com>
Gaetan Rivet [Thu, 3 Aug 2017 15:08:34 +0000 (17:08 +0200)]
net/failsafe: fix blank line parsing
When the output of an exec() slave definition is only a single newline
character, the fail-safe currently fails to parse the device with the
value returned by the rte_devargs library.
This behavior is incorrect, because the fail-safe should make a
difference between the absence of a device, and an erroneous device
declaration.
Fix the output sanitization in the case where no newline was at its end
and detect the special case of an absent device. The correct error code
is then returned.
When there is no preferred device, failsafe will always
try to scan for preferred device. And if there is no device
found with the exec option, popen() will get an empty output.
In this case, it was forgotten to close the file descriptor.
It is fixed by closing the file descriptor even if the output is empty.
Tiwei Bie [Tue, 1 Aug 2017 09:01:21 +0000 (17:01 +0800)]
vhost: make page logging atomic
Each dirty page logging operation should be atomic. But it's not
atomic in current implementation. So it's possible that some dirty
pages can't be logged successfully when different threads try to
log different pages into the same byte of the log buffer concurrently.
This patch fixes this issue.
Fixes: b171fad1ffa5 ("vhost: log used vring changes") Cc: stable@dpdk.org Reported-by: Xiao Wang <xiao.w.wang@intel.com> Signed-off-by: Tiwei Bie <tiwei.bie@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Gaetan Rivet [Thu, 3 Aug 2017 12:34:31 +0000 (14:34 +0200)]
eal: fix leak on hotplug parsing error
If rte_eal_devargs_parse fails, the rte_devargs has not yet been inserted
in the global list. When jumping to err_devarg, the removal fails and it
is not properly freed.
Free the allocated rte_devargs if its removal failed.
Gaetan Rivet [Wed, 2 Aug 2017 17:12:07 +0000 (19:12 +0200)]
devargs: fix policy check when adding arguments
Do not fail upon detecting a set bus configuration when adding an
rte_devargs. i.e. adding a blacklisted PCI device while the bus is
configured in whitelist mode.
This failure condition has been introduced when the bus policies
were implemented. This implementation however was meant to strictly
follow the existing API. This check was not and should not be performed
until the current API has been properly deprecated.
The same kind of check is already done when parsing EAL options.
Fixes: 02823c1db0bc ("devargs: parse bus policies") Signed-off-by: Gaetan Rivet <gaetan.rivet@6wind.com>
Gaetan Rivet [Wed, 2 Aug 2017 17:10:22 +0000 (19:10 +0200)]
eal: read and parse device option separately
When the EAL parses the common options given to the application,
not all subsystems are available. Some device drivers are registered
afterward upon dynamic plugin loading.
Devices using those drivers are thus unable to be parsed by any drivers
and are rejected.
Store the device options first and keep them for later processing.
Parse these right before initializing the buses, the drivers must have
been stabilized at this point.
Thomas Monjalon [Thu, 3 Aug 2017 13:53:50 +0000 (15:53 +0200)]
doc: notify libnuma dependency in release notes
Fixes: 066731939589 ("doc: add libnuma as dependency") Signed-off-by: Thomas Monjalon <thomas@monjalon.net> Acked-by: John McNamara <john.mcnamara@intel.com>
Harry van Haaren [Wed, 12 Jul 2017 13:38:52 +0000 (14:38 +0100)]
doc: add author on cc to git fixline alias
With this commit, the correct method to use git fixline to indicate
a fix of a previous commit has changed. The new rules require the
author of the original code that is being fixed to be on CC.
The logic behind this improvement is that if there is a genuine bug,
one of the ideal people to review is the author of the original code
being fixed. Adding them on Cc makes them aware of the patch, avoiding
it from being passed by accidentally while reading the mailing-list.
Given that the original author (now on Cc:) might not actually review,
there is no value in keeping the Cc: in git commit history. If the
original author performs a review, their Reviewed-by: or Acked-by: is
stored in git history (same as now).
Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com> Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com> Acked-by: John McNamara <john.mcnamara@intel.com> Acked-by: Thomas Monjalon <thomas@monjalon.net>
Thomas Monjalon [Sat, 15 Jul 2017 10:00:33 +0000 (12:00 +0200)]
devtools: add script to find duplicated includes
Based on Stephen's idea (originally implemented in a Perl script),
this is a shell script to find duplicated includes in a file.
It looks for all the .c and .h files of the git repository.
It is fast enough because automatically well parallelized.
Suggested-by: Stephen Hemminger <stephen@networkplumber.org> Signed-off-by: Thomas Monjalon <thomas@monjalon.net> Tested-by: Keith Wiles <keith.wiles@intel.com>
Thomas Monjalon [Mon, 31 Jul 2017 12:07:38 +0000 (14:07 +0200)]
devtools: ignore non merged tags for backport
When checking if a buggy commit was introduced in an old version,
the script compares last tag containing the bug and current version.
The non merged tags from non related branches must be ignored.
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
The Linux Getting Started Guide contains
parts which are specific for i40e PMD. This results
in confusion for users which read the guide at their
first try with DPDK.
Moving those parts to the i40e NIC manual.
Signed-off-by: Shahaf Shuler <shahafs@mellanox.com> Acked-by: John McNamara <john.mcnamara@intel.com>
git.droids-corp.org / dpdk.git / log