Age | Commit message (Collapse) | Author | Files | Lines |
|
It turns out that for scalar sizes 0..24, frames are always the same
size. That range includes all current use-cases - and then some - so
get rid of the hash table. Old code preserved under #ifdef
VLIB_SUPPORTS_ARBITRARY_SCALAR_SIZES.
Change-Id: Ic005c7143c9639f77d1a0fadd2fc0e90dccb68c1
Signed-off-by: Dave Barach <dbarach@cisco.com>
|
|
VPP supports two DPDK drivers for managing netvsc devices on
Azure/Hyper-V. The new netvsc PMD looks a lot like other PCI-based
PMDs but it requires recently added kernel support (>=4.17). The
older vdev_netvsc is an abstraction that manages the mlx4 VF
and tap device underlying the netvsc interface using the failsafe PMD.
Distros with older kernels (e.g. RHEL/CentOS 7.x) have to use vdev_netvsc.
At startup, netvsc devices are processed and an attempt is made to
initialize them for management by the netvsc PMD. If that fails, then
vlib_vmbus_bind_to_uio() returns early and the device can be initialized
for management by vdev_netvsc.
The operation that is supposed to fail if the netvsc PMD cannot be used
is registration of the netvsc device type ID with the uio_hv_generic
driver. This operation is attempted exactly once so it does not fail
for netvsc devices processed after the first one and they end up in a
state where they cannot be initialized for use by vdev_netvsc.
Only unset uio_new_id_needed if uio_hv_generic registration succeeds.
Change-Id: I6be925d422b87ed24e0f4611304cc3a6b07a34fd
Signed-off-by: Matthew Smith <mgsmith@netgate.com>
|
|
Remove logically dead code to keep coverity from complaining
Change-Id: If27d6684d19ab3c8886732a67922c86e5f0b3554
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
Change-Id: I550ef893e41d86310d4e5af16d6100b8e557b68a
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: I911b5070f544c3556ad1ca584d42689c00cc38f3
Signed-off-by: Filip Tehlar <ftehlar@cisco.com>
|
|
Comes in handy when looking at core files from optimized images.
Signed-off-by: Dave Barach <dave@barachs.net>
Change-Id: I45c8400d15614d5a04a3dcfb9b7daccab47ab446
|
|
Change-Id: I8b59b2e1c0525abf4b0492e50a7af57df4cd3ce2
Signed-off-by: Dave Barach <dave@barachs.net>
|
|
Change-Id: I6819dd9dbfc15c17740bdb98b51bdd639ef8c4d2
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: I500b8e72d7e021b9564dad0ffc087cb0e1ceb96f
Signed-off-by: Ole Troan <ot@cisco.com>
|
|
Moved code to the ethernet input node, and the interface output
path(s). Since we no longer skip ethernet-input, there's no reason
for device drivers to know anything about pcap rx tracing, etc.
Change-Id: I08d32fb1b90cbee1bd4f609837d533e047b36fa4
Signed-off-by: Dave Barach <dave@barachs.net>
|
|
Rewrite vlib_buffer_chain_linearize function so that it works as intended.
Linearize buffer chains coming out of reassembly to work around some
dpdk-tx issues. Note that this is not a complete workaround
as a sufficiently large packet will still cause the resulting chain to
be too long.
Drop features from reassembly code which relies on knowing which and how
many buffers were freed during linearization, buffer counts and tracing
capabilities for these cases.
Change-Id: Ic65de53ecb5c78cd96b178033f6a576ab4060ed1
Signed-off-by: Klement Sekera <ksekera@cisco.com>
|
|
Change-Id: I7f7a459f25d64ea5fa36e30d7dccc667bc19c5a9
Signed-off-by: Filip Tehlar <ftehlar@cisco.com>
|
|
Change-Id: I01c4f5755d579282773ac227b0bc24f8ddbb2bd1
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Some x86 CPUs have IOMMU capable dealing only with 39-bit address space
This patch also adds option to specify physmem base address from
startup.conf
Change-Id: I9e8abd26efb60e9c4ad54c035fb1751a4a61f4dc
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
The main thread squirrels away vlib_time_now (&vlib_global_main),
worker threads use it to calculate an offset in f64 seconds from their
own vlib_time_now(vm) value. We use that offset until the next barrier
sync.
Thanks to Damjan for the suggestion.
Change-Id: If56cdfe68e5ad8ac3b0d0fc885dc3ba556cd1215
Signed-off-by: Dave Barach <dave@barachs.net>
|
|
buffer_pool_index, in case of buffer_pool_index change,
should free buffers in queue before setting buffer_pool_index again
Change-Id: I846ead947a7b2d940c8fc747976239d608597391
Signed-off-by: Lollita Liu <lollita.liu@ericsson.com>
|
|
Change-Id: I4e836244409c98739a13092ee252542a2c5fe259
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Example:
buffers {
default data-size 1536
}
Change-Id: I5b4436850ca18025c9fdcfc7ed648c2c2732d660
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: Ib59a3c32af754a898ade17c42e60a88f48b797ff
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: I195c8eabc0ee67880f1e85fc7594b00be6b563e3
Signed-off-by: Dave Barach <dave@barachs.net>
|
|
It is causing compilation sloness with gcc-7 so removing it
before it was originally planned.
So far macros are left in the tree so we can know which nodes to
convert to new multiarch code.
Change-Id: Idb14622ca61fdce1eba59723b20d98715b7971e6
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: I8044b34a37fe1994a8dfa1ca89929f3642c72e8d
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Add a new command to dump vlib graph as graphviz/dot file
Change-Id: I43fc072cff8153ac500e5fbc6641a3705c2e995e
Signed-off-by: Benoît Ganne <bganne@cisco.com>
|
|
The assert allows easier debugging in gdb by aborting instead of exiting
cleanly. Also a core is generated so a stack trace can be obtained.
Change-Id: I2d05488c4522f4d4570fdfe0283130eb4c853d2a
Signed-off-by: Klement Sekera <ksekera@cisco.com>
|
|
Change-Id: I2cfb81b3c8809d027a3ee5f5f570668e60be08ef
Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com>
|
|
- DPDK overwrites metadata as part of rte_pktmbuf_init(...) so we need
reinitialize it
- additional checks added to ensure ref_count is never < 1
Change-Id: Ida336f81c4723e8f2e0ad4a70cb7b1ecfff978a0
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
This patch introduces following changes:
- deprecated free lists which are not used and not compatible
with external buffer managers (i.e. DPDK)
- introduces native support for per-numa buffer pools
- significantly improves performance of buffer alloc and free
Change-Id: I4a8e723ae47056717afd6cac0efe87cb731b5be7
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
This variable without any needed 'define' becoming unused.
Change-Id: I661a75a78dba03abb861ed918ad1e634a8ecd0af
Signed-off-by: Dmitry Vakhrushev <dmitry@netgate.com>
|
|
Change-Id: I3a33fb81f31ed473811e9e7a6197b81135913865
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: I79b213b34c6071d14acf1922f89037a4a5a36c45
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Add missing pre-input node runtime fork and refork code.
unix-epoll-input runs on all threads; each instance needs its own
runtime stats.
Change-Id: I16b02e42d0c95f863161176c4bb9f9917bef809d
Signed-off-by: Dave Barach <dave@barachs.net>
|
|
thanks, Eliot...
Change-Id: I8d8fee09bf1fe24933e6ef4e126dba8e22fe62b1
Signed-off-by: Dave Barach <dave@barachs.net>
|
|
As a FUD reduction measure, this patch implements 2-way parallel
counter collection. Synthetic stat component counter pairs run at the
same time. Running two counters (of any kind) at the same time
naturally reduces the aggregate time required by an approximate
factor-of-2, depending on whether an even or odd number of stats have
been requested.
I don't completely buy the argument that computing synthetic stats
such as instructions-per-clock will be inaccurate if component counter
values are collected sequentially. Given uniform traffic pattern, it
must make no difference.
As the collection interval increases, the difference between serial
and parallel component counter collection will approach zero, see also
the Central Limit theorem.
Change-Id: I36ebdcf125e8882cca8a1929ec58f17fba1ad8f1
Signed-off-by: Dave Barach <dave@barachs.net>
|
|
Change-Id: I9df96264e30806ac3daf7121f314f34f06232413
Signed-off-by: Damjan Marion <damjan.marion@gmail.com>
|
|
Change-Id: I1c12e2941cae198ededbb65eb5be51a4eabe2c1b
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: I0ba5175be077c40556f2a3ce629c5bbcd71e0a81
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: I5fe01e918b534d1ac1d47f0d7610536f45c3158c
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: Ia083050389853c25b069f0f8286d50d3f4aef527
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: I3bb1d9f83dd08f4b93acd4a281bfec0674e39c2e
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: If88ccd965122b9318a39a8d71b53334cd1fd81e4
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
1. Otherwise, the log classes with same prefix, e.g. "abc" and
"abc-de" will all be registered into log class "abc";
2. Minor improvement for test log help string.
Change-Id: I8d93be5e8fa67db6012198b3442a9e2bddcb744a
Signed-off-by: Su Wang <su.z.wang@ericsson.com>
|
|
For some reason, GCC 8 in debian is pickier than GCC 8 in ubuntu. It complains
about things in strncpy like this
/home/sluong/vpp/src/vlib/linux/pci.c:485:7: error: ‘strncpy’ output may be
truncated copying 15 bytes from a string of length 255 [-Werror=stringop-truncation]
strncpy (ifr.ifr_name, e->d_name, sizeof (ifr.ifr_name) - 1);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/sluong/vpp/src/vlib/linux/pci.c: At top level:
It also complains similar things in string_test.c
The fix in pci.c is to convert strncpy to use clib_strncpy
The fix in string_test.c is condiational compile the complained code for GCC 8.
Change-Id: Ic9341ca54ed7407210502197a28283bc42c26662
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
Change-Id: I6048c6a51efa826ac333f7d15919cb87dd766d74
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: Idb224b807373128b78dc46f177b2d2e0ed288349
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
This reverts commit 1e59f9ddbdda14591967e1d66eab8623f9ba58e4.
Change-Id: Iae1d372b887e170d28cac2fe4c61325ee5a5894a
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Storing buffer in local template seems to be better option....
Change-Id: I1a2fdd68cb956f99a5b36d2cd810fc623e089bcf
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
The code that was manipulating interface names with ifreq was
causing warnings about possible truncation and non terminated
strings.
These are warnings only since kernel would allow a interface
name > 15 characters anyway.
Change-Id: I794a94fe310b8568403d4e3523c61d53468a6f02
Reported-by: Burt Silverman <burtms@gmail.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
|
|
Doing strncpy(ifr.ifr_name, s, sizeof(ifr.ifr_name)) will cause
a warning about string truncation with GCC 8 (and other tools).
Fix this by using sizeof(ifr.ifr_name) - 1. Also, there is no
need to manually zero the end of the string since the whole
ifr structure is already zeroed by memset.
Change-Id: I9440d602ecdd9f8592b69bab2e77479146d00d76
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
|
|
Change-Id: Ia34a4278eedc8cf450688b1fa0291e1f976868d3
Signed-off-by: Dave Barach <dave@barachs.net>
|
|
Change-Id: Ic56ee4ce83b282a5f0f5aed500721fe639b941b3
Signed-off-by: Damjan Marion <damarion@cisco.com>
|