Crash was due to loopDelay being 0 in case of topSpeed. For ttags, we
recalculatue loopDelay based on max line rate and avg pkt size.
The problem doesn't exist in interleaved mode because minGap and duration
are non-zero in the top speed tx case.
Tx/Rx stream stats related threads no longer have a direct reference to the
port's stream stats - instead they have their own copy that they keep and
return (in a reset-on-read fashion) when asked for. Each copy also has it's
own lock to prevent contention for read/update/clear.
PcapPort now fetches Tx(Transmitter) and Rx(Poller) stats on demand and
updates the port's stream stats - under protection of a lock.
We are not rewriting L4Cksum for ttag packets at the moment. See
comment in packetsequence.h
Commenting out this code doesn't seem to observably improve tx
performance though.
The latency code seems to reduce stream stats max tx performance
by around 3 to 5%. Recovering this may be done separately as part
of overall performance optimzation of Tx code.
The interleaved mode's single packet set MUST always have 0 delay
for accurate rate.
Before latency code, the interleaved packet set was added implicitly
and had 0 delay. Latency code added explicit packet set and used
avgDelay for the set. The avgDelay was needed for the original algo
used to determine when to send ttag packets. That algo is no longer
used (ttag markers are now explicitly configured by AbstractPort).
Turbo still needs avgDelay for other use, but changes will be made in
Turbo code for that.
There are 2 changes -
1. Encode txPort in ttag packets and use it at TxTtagStats and
RxPcapStats to identify Tx and Rx packets respectively
2. Don't use pcap_sendqueue_transmit() if stream timing is in use -
since we can't modify TTAG packets inside that API
In a previous commit, we start/stop these timers based on number of
ports tracking stream stats triggered by RPCs. However, timers cannot
be triggered across threads (RPC thread and main thread in this case).
This fix uses a queued connection to post it to the other queue.
If and when we remove PacketSequence::ttagL4ChecksumOffset we will take a call
if we should revert back to passing seq->sendQueue instead of seq at that time
The stream timingHash is read by getStreamStats() while it is read/write
for processRecords(), the latter is a more frequent operation so there's
no real benefit of using a read-write lock instead of simple mutex.
PcaptxTtagStats inherits from PcapSession which already includes a protected
handle_ member.
This removal was likely left off when PcapTxTtagStats started inheriting from
PcapSession.
The previous commit changed the algo to determine which packets were Ttag'd,
but changes were done only for interleaved mode.
This commit adds the changes required for sequential mode.
The algo works for the following cases of interleaved streams -
* pktListDuration < ttagTimeInterval
* pktListDuration > ttagTimeInterval
* some streams have Ttag, some don't
- first stream has Ttag
- first stream does NOT have Ttag
* no streams have Ttag
Changes for sequential mode are pending
Interleaved mode used an implicitly added packet set in both base and Turbo
code. This has been chaned to use an explicit mode to keep things consistent.
Turbo code still has the implicit packet set related code - that needs to be
removed, once the explicit packet set code is validated and tested.
This singleton class will keep track of Ttag timing across all ports and GUIDs.
A bunch of FIXMEs/TODOs are pending for this class implementation; also this
class has not been hooked up to the rest of the code yet.
For now we are just debug printing timestamp with T-TagId and GUID. We
need to store this tuple and compare when we Rx the same - this will be
in a upcoming commit.
As part of Turbo changes, we made changes to create explicit packet
sets, but for the base code we continued creating implicit packet
sets for some cases. With this change we don't create any implicit
packet set.
This change needs to be tested thoroughly for multiple cases.
