[google/syzkaller] b541be: dashboard/config/linux: kcsan: remove KCSAN_DEBUG

[Marco Elver]

Branch: refs/heads/master
Home: https://github.com/google/syzkaller
Commit: b541beac5b929237d926ac9a616aad960822744a
https://github.com/google/syzkaller/commit/b541beac5b929237d926ac9a616aad960822744a
Author: Marco Elver <el...@google.com>
Date: 2021-10-20 (Wed, 20 Oct 2021)

Changed paths:
M dashboard/config/linux/bits/kcsan.yml

Log Message:
-----------
dashboard/config/linux: kcsan: remove KCSAN_DEBUG

CONFIG_KCSAN_DEBUG no longer exists, remove it. Older kernels will set
it to 'n' by default.


Commit: 418a00eb23ef6f6d2f60255c6185520ceaa2d0db
https://github.com/google/syzkaller/commit/418a00eb23ef6f6d2f60255c6185520ceaa2d0db
Author: Marco Elver <el...@google.com>
Date: 2021-10-20 (Wed, 20 Oct 2021)

Changed paths:
M dashboard/config/linux/bits/kcsan.yml
M dashboard/config/linux/upstream-kcsan.config

Log Message:
-----------
dashboard/config/linux: kcsan: do not pretend writes are atomic

Do not pretend writes are atomic. While this may currently be a
prevailing preference in the community, it does hide more interesting
bugs [1].

Since we moderate data races anyway, and are still drowning in data
races, a few more won't hurt.

What it does help with, however, is getting better signals about harmful
data races. Most of the read/write data races provide weak signals, but
write/write data races should provide a stronger harmfulness signal [1],
which will help us in selecting data races to investigate further.

[1] https://googleprojectzero.blogspot.com/2021/10/how-simple-linux-kernel-memory.html


Compare: https://github.com/google/syzkaller/compare/466b7db1ce8f...418a00eb23ef