https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=283903
--- Comment #33 from Guillaume Outters <guillaume-free...@outters.eu> ---
(In reply to Bjoern A. Zeeb from comment #32)
IT WORKS!
Thank you very much for having taken the time to review the whole code, putting
the missing spin_locks where necessary. It reminds me those good old times of
courses about mutex locking…
-----------------
My test procedure
On kernels:
- I (14.1-RELEASE + my printf traces A, B, C, D of comment #29, systematic (not
conditioned to any compat.linuxkpi.skb.debug bit, thus outputing since boot))
- II (I + your D49101 patch, hand-reworked to minimize changes in order not to
conflict with my own traces (removing of hunks of only comments, const,
variables rename as head -> q, and translation of functions (instead of moving
them I let them in place but added the necessary declaration)))
For each test I did:
# boot
# wayfire
# foot terminal:
sudo sysctl compat.linuxkpi.skb.debug=0x10
scp /tmp/1 mb:/tmp/1 & # My 10 MiB file born from /dev/random
scp mb:/tmp/1 /tmp/2 & # 10 MiB too, as the file is already in place and the
previous scp works inplace.
date
while true ; do vmstat -m | grep skb | tr \\012 \\011 ; date ; sleep 5 ; done
Tests:
- ✗ I: /!\ vmstat -m | grep skb instantly grew up
(from 16 MB to 100 MB at the time the scp finished)
- ✓ II: stable even after the scps
- ✗ I: /!\ freeze
- ✓ II: stable
- ✗ I: /!\ vmstat -m instantly grew up
- ✓ II: stable
(in fact I'm browsing to write this comment
while still not having unset compat.linuxkpi.skb.debug)
--
You are receiving this mail because:
You are on the CC list for the bug.
