Although ppc supports both endianesses, qemu supposes that the cpu is
big endian and enforces this for the virtio-net stuff.
Fix PCI accesses in legacy mode. Only ppc64le is supported at the moment.
Signed-off-by: David Marchand <david.marchand@6wind.com>
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
On PPC64, the ioports are mapped in memory. Implement the missing part
of ioport API for PPC64 when using uio. This may also work on other
architectures but it has not been tested.
Signed-off-by: David Marchand <david.marchand@6wind.com>
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Split pci_parse_sysfs_resource() and introduce
pci_parse_one_sysfs_resource() that parses one line of sysfs resource
file.
This new function will be exported and used in next commits when
mapping the ioports resources.
No functional change.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
In a previous commit, the file used to map the PCI resources changed
from "/dev/uio<x>" to "/sys/bus/pci/devices/<busaddr>/resource", making
the comment wrong. Remove it.
Fixes: 9e67561acd ("eal/linux: mmap uio resources using resourceX files")
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
From iopl(2) man page: "This call is mostly for the x86 architecture. On
many other architectures it does not exist or will always return an
error".
This patch removes the call to iopl() in rte_eal_iopl_init() for
architectures other than x86, and always return 0 (success). This was
already done for ARM in
commit 0291476ae3 ("eal/linux: never check iopl for arm")
Next patches will introduce the support of memory mapped IO resources
for architectures != x86.
On BSD, there is nothing to do as open("/dev/io") already does the
proper thing. See man IO(4).
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Fix some typos and add missing comments related to ioports API in
rte_pci.h.
Fixes: 756ce64b1 ("eal: introduce PCI ioport API")
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
The behavior of PKT_RX_VLAN_PKT was not very well defined, resulting in
PMDs not advertising the same flags in similar conditions.
Following discussion in [1], introduce 2 new flags PKT_RX_VLAN_STRIPPED
and PKT_RX_QINQ_STRIPPED that are better defined:
PKT_RX_VLAN_STRIPPED: a vlan has been stripped by the hardware and its
tci is saved in mbuf->vlan_tci. This can only happen if vlan stripping
is enabled in the RX configuration of the PMD.
For now, the old flag PKT_RX_VLAN_PKT is kept but marked as deprecated.
It should be removed from applications and PMDs in a future revision.
This patch also updates the drivers. For PKT_RX_VLAN_PKT:
- e1000, enic, i40e, mlx5, nfp, vmxnet3: done, PKT_RX_VLAN_PKT already
had the same meaning than PKT_RX_VLAN_STRIPPED, minor update is
required.
- fm10k: done, PKT_RX_VLAN_PKT already had the same meaning than
PKT_RX_VLAN_STRIPPED, and vlan stripping is always enabled on fm10k.
- ixgbe: modification done (vector and normal), the old flag was set
when a vlan was recognized, even if vlan stripping was disabled.
- the other drivers do not support vlan stripping.
For PKT_RX_QINQ_PKT, it was only supported on i40e, and the behavior was
already correct, so we can reuse the same bit value for
PKT_RX_QINQ_STRIPPED.
[1] http://dpdk.org/ml/archives/dev/2016-April/037837.html,
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Following the discussion on dpdk-users [1], enhance the API
documentation of rte_eth_tx_burst() to specify that the mbufs have to be
allocated from a pool.
[1] http://dpdk.org/ml/archives/users/2016-June/000618.html
Reported-by: Xiaoban Wu <xiaoban_wu@student.uml.edu>
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
When starting the ip_fragmentation or ip_reassembly example
on several sockets, it fails.
The name of the lpm6 table is the same on every socket,
resulting in a table creation failure (-EEXIST). The failure
appeared after:
commit f82f705b63 ("lpm: fix allocation of an existing object")
Indeed, before this commit the returned value when the existing
table, which was probably a bug in that case: one table for 2
sockets for lpm6, and one per socket for lpm.
Fixes: 74de12b7b6 ("examples/ip_fragmentation: overhaul")
Fixes: b84fb4cb88 ("examples/ip_reassembly: overhaul")
Reported-by: Heng Ding <hengx.ding@intel.com>
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
This patch fixes rte_memcpy performance in Haswell and Broadwell for
vhost when copy size larger than 256 bytes.
It is observed that for large copies like 1024/1518 ones, rte_memcpy
suffers high ratio of store buffer full issue which causes pipeline
to stall in scenarios like vhost enqueue. This can be alleviated by
adjusting instruction layout. Note that this issue may not be visible
in micro test.
How to reproduce?
PHY-VM-PHY using vhost/virtio or vhost/virtio loop back, with large
packets like 1024/1518 bytes ones. Make sure packet generation rate
is not the bottleneck if PHY-VM-PHY is used.
Test report: http://dpdk.org/ml/archives/dev/2016-May/039716.html
Signed-off-by: Zhihong Wang <zhihong.wang@intel.com>
Tested-by: Qian Xu <qian.q.xu@intel.com>
The mempool local cache was not initialized properly leading to
undefined behavior in cases where the allocated memory was used
previously and left with data.
Fixes: 213af31e09 ("mempool: reduce structure size if no cache needed")
Signed-off-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
There are 2 new fake devices for testing PCI infra. All the fake devices
are now identified by non-existing vendor and device IDs so there is no
real driver to bind to them. The testing drivers match those IDs.
Suggested-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
This patch defines the target and arch value of objcopy program for
IBM POWER PPC64 little endian architecture.
Fixes: 99d6231fdc ("mk: define objcopy-specific target and arch")
Signed-off-by: Chao Zhu <chaozhu@linux.vnet.ibm.com>
This script parses the application configuration file and detects all the
pipelines specified therein, and then, it generates all the possible mappings
of those pipelines on the specified CPU core-list.
As a result, each of the possible pipeline-to-core mappings is saved as
separate output configuration file. For example- if input file is
edge_router_downstream.cfg with 3 pipeline (excluding pipeline 0) and
core-list is “1, 2”, following combinations will be generated-
Pipeline 1 Pipeline 2 Pipeline 3
Core = 1 Core = 1 Core = 2
Core = 1 Core = 2 Core = 1
Core = 2 Core = 1 Core = 1
Core = 2 Core = 2 Core = 1
Core = 2 Core = 1 Core = 2
Core = 1 Core = 2 Core = 2
Core = C1 Core = C1H Core = C2
Core = C1 Core = C2 Core = C1H
Core = C2 Core = C1 Core = C1H
This script will help users to analyse the performance of application by
evaluating all the generated configuration files with different
pipelines-to-core mappings and obtaining the application configuration file
with best performance.
To run the script, issue the following command;
./pipeline-to-core-mapping.py -i <input_configuration_file> -pc "s<socket-id>c<core-id>"
Some optional arguments are as follows:
-h, --help show this help message and exit
-ht {ON,OFF}, --hyper-thread {ON,OFF}
enable/disable hyper threading. default is ON
-nO, --no-output-file
disable output config file generation. Output file
generation is enabled by default
Signed-off-by: Guruprasad Mukundarao <guruprasadx.rao@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
In ip_pipeline app, the structure app_thread_data needs to be aligned to
the cache line boundary as threads on different cpu cores are accessing
fields of the app->thread_data and having this structure not aligned on
cacheline boundary leads to false cacheline sharing.
Fixes: 7f64b9c004 ("examples/ip_pipeline: rework config file syntax")
Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
Return value of function app_pipeline_type_find is not checking before
dereference. Fix this problem by adding checking condition.
Coverity issue: 127196
Fixes: b4aee0fb9c ("examples/ip_pipeline: reconfigure thread binding dynamically")
Signed-off-by: Marcin Kerlin <marcinx.kerlin@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
Function like 'rte_kni_rx_burst()' keeps
allocating 'MAX_MBUF_BURST_NUM' mbufs to
kni fifo queue unless the queue's capacity
('KNI_FIFO_COUNT_MAX') is reached. So, if
the mempool is under-provisioned, user may
run into "Out of Memory" logs from KNI code.
This commit documents the need to provision
mempool capacity of more than
"2 x KNI_FIFO_COUNT_MAX" for each KNI interface.
Signed-off-by: Alex Wang <alex@awakenetworks.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
The 'mbufs' alloc/free descriptions for
'rte_kni_tx_burst()' and 'rte_kni_rx_burst()'
should be inverted.
Fixes: 3fc5ca2 ("kni: initial import")
Signed-off-by: Alex Wang <alex@awakenetworks.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
This patch docs the issue on EAL argument that the last EAL
argument is replaced by program name in argv[].
Reported-by: Ziye Yang <ziye.yang@intel.com>
Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
This patch is used to add the class_id (class_code,
subclass_code, programming_interface) support for
pci_device probe. With this patch, it will be
flexible for users to probe a class of devices
by class_id.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
The commit 66819e6 has introduced a dependency on libarchive to be able
to use some tar resources in the unit tests.
It is now an optional dependency because some systems do not have it
installed.
If CONFIG_RTE_APP_TEST_RESOURCE_TAR is disabled, the PCI test will not
be run. When a "configure" script will be integrated, the libarchive
availability could be checked to automatically enable the option.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Reviewed-by: Jan Viktorin <viktorin@rehivetech.com>
Recent CPU's can easily wrap around a 32-bit unsigned int in
the mempool perf test. Increase to a 64-bit uint.
Signed-off-by: David Hunt <david.hunt@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
Bond device name was too long (grather than 32 signs) that
cause mempool allocation to fail.
Fixes: 92073ef961 ("bond: unit tests")
Signed-off-by: Michal Jastrzebski <michalx.k.jastrzebski@intel.com>
Acked-by: Bernard Iremonger <bernard.iremonger@intel.com>
Tested-by: Huilong Xu <huilongx.xu@intel.com>
app/test/test_cryptodev.c:
In function ‘create_snow3g_cipher_operation_oop.clone.15’:
x86_64-native-linuxapp-gcc/include/rte_memcpy.h:796:14
error: array subscript is above array bounds.
In test_cryptodev.c:
2429 rte_memcpy(sym_op->cipher.iv.data, iv, iv_len);
When iv_len is declared as 'unsigned int', rte_memcpy evaluates code for
buffer size bigger than 255, but while 'iv' array is 64 bytes long, it
causes 'above array bounds' warning in gcc 4.5 and breaks compilation.
Using uint8_t as a size of copied block prevents to evaluate in rte_memcpy
code for length bigger than 255, causing the problem.
The root of this issue and solution is the same as for commit 2c007ea106
("app/test: fix array overflow warning with gcc 4.5")
Fixes: 9727af14b0 ("app/test: add out-of-place symmetric crypto operations")
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
'red_autotest' changed to run only functional tests without test #4 which was
taking ~53 seconds. 'red_autotest' takes ~2[s] now.
'red_perf' has been added to run performance tests only).
'red_all' has been added to run all functional tests (including #4) and
perfromance tests. This reflects current 'red_autotest' behavior.
Other changes:
- machine TSC clock frequency detection takes place only once now.
- timeouts and number of iterations in functional tests have been reduced
in order to shorten test duration.
Signed-off-by: Tomasz Kantecki <tomasz.kantecki@intel.com>
Dumping of devices in a unittest is useless. Instead, test whether
the test has been set up well - i.e. there are no devices.
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
The SYSFS_PCI_DEVICES is a constant that makes the PCI testing
difficult as it points to an absolute path. We remove using this
constant and introducing a function pci_get_sysfs_path that gives
the same value. However, the user can pass a SYSFS_PCI_DEVICES env
variable to override the path. It is now possible to create a fake
sysfs hierarchy for testing.
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
The current test_pci is just a single test case that tests the
blacklisting of devices. Rename it to test_pci_blacklist and call it
from the test_pci. The setup and cleanup are moved out of the
test_pci_blacklist entirely to cover all other tests.
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
The test unregisters all drivers before start. The drivers were stored
into a fixed-sized array. This is inflexible. This patch change this to
utilize a linked list for the same purpose.
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
When a more complex resource (a file hierarchy) is needed, packing
every single file as a single resource would be very ineffective. For
that purpose, it is possible to pack the files into a tar archive,
extract it before test from the resource and finally clean up all the
created files.
This patch introduces functions resource_untar and resource_rm_by_tar
to perform those tasks. An example of using those functions is included
as a test.
A new dependency is required to build the app/test: libarchive.
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
A resource can be written into the target filesystem by calling
resource_fwrite or resource_fwrite_file. Such file can be created
before a test is started and removed after the test finishes.
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
To include resources from other source that the C source code we
can take advantage of the objcopy behaviour, i.e. packing of an
arbitrary file as an object file that is linked to the target program.
A linked object file is always accessible as a pair
extern const char beg_<name>;
extern const char end_<name>;
(extern const char siz_<name>;)
A unit test that packs the resource.c source file is included.
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
Certain internal mechanisms of DPDK access different file system
structures (e.g. /sys/bus/pci/devices). It is difficult to test
those cases automatically by a unit test when such path is not
hard-coded and there is no simple way how to distribute fake ones
with the current testing environment.
This patch adds a possibility to declare a resource embedded in
the test binary itself. The structure resource cover the generic
situation - it provides a name for lookup and pointers to the
embedded data blob. A resource is registered in a constructor by
the macro REGISTER_RESOURCE.
Some initial tests of simple resources is included and added into
the group_1.
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
The program objcopy uses non-standard conventions to name the
target and arch. Define the values for supported architectures
(tile and ppc_64 are missing).
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
This patch adds missing DEPDIRS to avoid any library referring to
symbols they are not linked against.
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
Add the missing external dependency to pthread to avoid referring to
symbols the library is not linked against.
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
The --whole-archive argument is only required for plugins (drivers)
and libraries used by these plugins.
Currently it covers all libraries.
Reducing the scope of this argument slightly reduce final application size
when statically linked.
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
As stated in the comment:
Order is important: from higher level to lower level
This is an attempt to make the layering order better respected.
It will help to restrict the --whole-archive scope for plugins.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Replace --no-as-needed linker flag with --as-needed flag, which will
only link libraries directly called by application.
It can be achieved now that the libraries dependencies are handled
properly.
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
>From "man ld":
Using this option has a significant performance cost.
It is best to use it only when there are unavoidable
circular references between two or more archives.
Remove the option since it does not seem necessary.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
On a linker command line, the dependencies must be declared after
the libraries using them.
It will avoid some issues when building an application with static
libraries and --as-needed option.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
--exclude became a positional option in tar 1.29, breaking the
test app filtering in "make install", causing .map files and all test
apps to get installed in bindir. Adjust the tar arguments accordingly,
this is compatible with older versions too since they do not care about
the order.
Ref: https://bugzilla.redhat.com/show_bug.cgi?id=1337864
Fixes: 6b62a72a70 ("mk: install a standard cutomizable tree")
Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
The libraries rte_mempool and rte_ring are not used in EAL,
except for the ivshmem part (CONFIG_RTE_LIBRTE_IVSHMEM).
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: David Marchand <david.marchand@6wind.com>
The log history uses rte_mempool. In order to remove the mempool
dependency in EAL (and improve the build), this feature is deprecated.
The ABI is kept but the behaviour is now voided because it seems this
function was not used. The history can be read from syslog.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: David Marchand <david.marchand@6wind.com>
Up to now dependencies between DPDK internal libraries have been
untracked at shared library level, requiring applications to know
about library internal dependencies and often consequently overlinking.
Since the dependencies are already recorded for build ordering in the
makefiles with DEPDIRS-y we can use that information to generate LDLIBS
entries for internal libraries automatically.
Also revert commit 8180554d82 ("vhost: fix linkage of driver with
library") which is made redundant by this change.
Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
Acked-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
