Hi Yalei, I guess there’s two solutions to that. Set min_chunk to something like clib_min (16 << 10, fifo_size/2) or remove min_chunk entirely.
I’m fine with either of those. Feel free to push a patch! Thanks, Florin > On Aug 8, 2018, at 12:17 AM, 汪亚雷 <wyland...@gmail.com> wrote: > > Hi Florin, > > These days I tested the tcp_echo of master branch, and found an issue: when I > set the fifo-size in client to value that lower then 16, it test will hang > because nothing send to server and of cause no recv too. > > After some investigation, it is related to this lines in send_test_chunk. > > ----- > if (enq_space < clib_min (bytes_this_chunk, min_chunk)) > return; > ------ > > From the code logic, it may want to use the available space in fifo, but > there is a hard code length 16<<10 for min_chunk, so looks like we could > either remove > this line( because code still try to get the available length of fifo in > following code), or > add limitation announcement for fifo-size configuration. > > > /Yalei
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#10067): https://lists.fd.io/g/vpp-dev/message/10067 Mute This Topic: https://lists.fd.io/mt/24228320/21656 Group Owner: vpp-dev+ow...@lists.fd.io Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-