Age | Commit message (Collapse) | Author | Files | Lines |
|
Type: refactor
This patch refactors the offload flags in vlib_buffer_t.
There are two main reasons behind this refactoring.
First, offload flags are insufficient to represent outer
and inner headers offloads. Second, room for these flags
in first cacheline of vlib_buffer_t is also limited.
This patch introduces a generic offload flag in first
cacheline. And detailed offload flags in 2nd cacheline
of the structure for performance optimization.
Change-Id: Icc363a142fb9208ec7113ab5bbfc8230181f6004
Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com>
|
|
Type: improvement
Change-Id: I4008cadfd5141f921afbdc09a3ebcd1dcf88eb29
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
When using classifier to filter traces, not all packets will be traced.
In that case, we should only count traced packets.
Type: fix
Change-Id: I87d1e217b580ebff8c6ade7860eb43950420ae78
Signed-off-by: Benoît Ganne <bganne@cisco.com>
|
|
outbound:
wrong header len computation
gso size and header length need to be set in the first segment of the
chain
inbound:
EOP may have zero length descriptor to terminate the chain
missing endian conversion for ethertype
Type: fix
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: Iaa003c0e9af3ead4df6c6c0d5772a179d2ff15c4
|
|
Type: feature
Change-Id: I913f08383ee1c24d610c3d2aac07cef402570e2c
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
vlib_increment_combined_counter takes sw_if_index, not hw_if_index. Using
hw_if_index may work as long as there is no subinterface created to cause
hw_if_index and sw_if_index to differ.
Type: fix
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: If50412dc1e84c4f1f5b977b58d0e1aeb5ab8ebe6
|
|
gso option for vmxnet3 is per VPP process currently. There is no reason why
we cannot provide per interface gso support.
Type: feature
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: Ife962b52221191050dedd18252b859880ccd7599
|
|
Queue everything to ethernet_input to not break pcap rx trace.
Fix a minor typo in vmxnet3_api.c
Type: fix
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: I7df7b3e20b525e3d6f7421c8e213b6541cdd15b6
|
|
vnet_buffer2 -> gso_l4_hdr_sz is not set correctly because
vnet_buffer->l4_hdr_offset is not set yet at the top. Move the code to set
vnet_buffer->gso_l4_hdr_sz further down.
Change-Id: I0110e4687572bb93cb593d7d4d14a3d5083af13a
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
Add lro/tso and rx checksum support. lro/tso is configured via startup.conf
vmxnet3 { lro }
It is disable by default due to not all versions of ESXi supports it.
Change-Id: Icf224ff528884ecd9e655b4fcf4481194e8c5a63
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
In output.c, we buffer the descriptors and call vmxnet3_reg_write_inline
once outside the loop. This change improves the performance dramatically.
When refilling the ring, there is no need to inform the device unless
explicitly specified by the device (ctrl.update_prod == 1)
Change-Id: I7031d58bff0d249e913d14236d416c91eb6ab94a
Signed-off-by: Steven <sluong@cisco.com>
|
|
try harder on output - if there is no descriptor space available, try to free
up some and check again.
make sure we free the buffer if error is encoutered on input.
Change-Id: I41a45213e29de71935afe707889e515037cd081f
Signed-off-by: Steven <sluong@cisco.com>
|
|
Coverity complains that hb might be null when !prev_b0 is true. It may be right, but not likely.
To make the code more explicit when it parses a complete packet, set got_packet = 1 and test
got_packet instead of prev_b0 which is somewhat obscure.
Use vlib_buffer_free_one rather than vlib_buffer_free_no_next in output.c which takes care of
chain buffers.
Change-Id: I6e0587605fdc622c3f55fedd74c474666631ed91
Signed-off-by: Steven <sluong@cisco.com>
|
|
outbound -- if we have to drop the packet due to no descriptor space is available, drop the
whole packet, not fragments.
inbound -- check and drop error packets
Change-Id: Ida1d32e61521bafd67f714d729ad53cd7c487dc6
Signed-off-by: Steven <sluong@cisco.com>
|
|
Under extreme condition, it is possible for input node to have more than 256 packets available
to process. Add a check to the while loop to prevent overrun buffer_indices and nexts.
Change-Id: Id9830e5c254f04216eb79e864a5460e08cf2e4e5
Signed-off-by: Steven <sluong@cisco.com>
|
|
chain buffers
If we invoke the subject API with 2 buffers, each with 2 chains, the content of
buffers[0] should be head of 1st buffer, and content of buffer[1] should be
head of 2nd buffer.
What the code did was to put
buffers[0] = head of 1st buffer
buffers[1] = next chain of 1st buffer
buffers[2] = head of 2nd buffer
buffers[3] = next chain of 2nd buffer
This is wrong and can cause crash.
The fix is to only increment bi when the entire packet is found which may consist
of multiple fragments.
Change-Id: If02cd5186a039d7a6c08a0959112840bdb242317
Signed-off-by: Steven <sluong@cisco.com>
|
|
Support feature arc
Support set interface next node
Enqueue packet to either ip4, ip6, or ethernet_input based on packet type
Change-Id: I11fb6bba98ee0bfa1d057474fc6cf3be48ef1c48
Signed-off-by: Steven <sluong@cisco.com>
|
|
Implemented vmxnet3 deivice driver for VMWare ESXi. Tested with Ubuntu 18.04
connected to ESXi 6.0
Ubuntu-18.04 (VPP) --- ESXi-6.0
Change-Id: I85fbc86f2d8532b017bc4271612d17e24e498e4d
Signed-off-by: Steven Luong <sluong@cisco.com>
|