Petr Vyazovik
Petr Vyazovik
` 1.24% ksoftirqd/2 [kernel.vmlinux] [k] rb_erase ` here is most likely related to out-of-order internal TCP queue processing, not HTTP/2 streams processing. ``` Samples: 262K of event 'cycles', Event count...
- [x] Some strange crash on `if-non-match` parsing: ``` [ 362.596229] [tempesta fw] tfw_h2_parse_req_hdr: fin=1, len=15, data="6331593c-3cd1", req=[00000000529c3f7d] [ 362.596236] [tempesta fw] enter FSM at state 0000000000000000 [ 362.596245] [tempesta...
Unfortunately, i wasn't able to reproduce the following issues: - [ ] [Warning in `tfw_hpack_encode()`](https://github.com/tempesta-tech/tempesta/issues/1387#issuecomment-594539715) - [ ] [Warning in `tfw_huffman_decode()`](https://github.com/tempesta-tech/tempesta/issues/1387#issue-575394510) Task №5
I believe the initial problem, which was: `BUG tfw_stream_cache (Tainted: G B W O ): Objects remaining in tfw_stream_cache on __kmem_cache_shutdown()` has already been fixed by [7a00fdf05b3ae65ed5861ce46c44f2f1e300deb3](https://github.com/tempesta-tech/tempesta/commit/7a00fdf05b3ae65ed5861ce46c44f2f1e300deb3). Since we don't...
I've rechecked all the warnings on TFW master branch from [12 Jan 2021 ](https://github.com/tempesta-tech/tempesta/commit/e593ddc6c59a8a03b911399d5ce11b1979bc8883) using tempesta-test [de0cbb8](https://github.com/tempesta-tech/tempesta-test/commit/de0cbb8) and then ran the same tests on [#1777](https://github.com/tempesta-tech/tempesta/pull/1777). I don't see any warnings/crashes...
Regarding this bug: ``` [ 350.787612] skbuff: skb_under_panic: text:ffffffff89212c3c len:64 put:20 head:ffff904deb8c1e00 data:ffff904deb8c1df0 tail:0x30 end:0xc0 dev: [ 350.787981] ------------[ cut here ]------------ [ 350.788284] kernel BUG at net/core/skbuff.c:111! ``` The...
The issue was originally reproduced in a setup when both Tempesta and backend (nginx) are on the same machine and communication is being performed through loopback interface with MTU ==...
So almost all the problems reported here are caused by a lack of `headroom` space, as i've explained in the previous comment. I don't see some warnings, that were reported...
I believe this [problem](https://github.com/tempesta-tech/tempesta/issues/1703#issuecomment-1308472743) is now fixed by [PR 1837](https://github.com/tempesta-tech/tempesta/pull/1837) by @const-t