[tarantool-patches] [rfc 3/4] core/fiber: Put static watermark into stack to track its usage
Konstantin Osipov
kostja at tarantool.org
Tue Mar 5 11:20:29 MSK 2019
* Cyrill Gorcunov <gorcunov at gmail.com> [19/03/03 23:25]:
> We want to detect a situation where task in fiber is too eager for
> stack and close to its exhausting. For this sake upon stack creation
> we put 8 marks on last stack page with step of 128 bytes. Such params
> allows us to fill ~1/4 of a page, which does seem reasonable but
> we might change this params with time.
>
> Since the watermark position is permanent and some task is close
> to stack limit we print about the situation once to not spam
> a user much and stop putting the mark on recycling.
>
> Closes #3418
As a policy we write meaningful comments for every function, static or not,
obvious or not. The idea of such a comment is to reflect the
author's intent and technical trade-offs considered when designing
a function. The comment should preferably be concise, however.
> +/*
> + * Random values generated with uuid.
> + * Try to fit a cache line.
I don't understand this comment. Since you're putting a poison-gap
between each number, how could you possibly fit a cache line?
> + */
> +static const uint64_t poison_pool[] = {
> + 0x74f31d37285c4c37, 0xb10269a05bf10c29,
> + 0x0994d845bd284e0f, 0x9ffd4f7129c184df,
> + 0x357151e6711c4415, 0x8c5e5f41aafe6f28,
> + 0x6917dd79e78049d5, 0xba61957c65ca2465,
> +};
> +
> +/*
> + * We poison by 8 bytes as it natural for stack
> + * step on x86-64. Also 128 byte gap between
> + * poison values should cover a common cases.
> + /** Stack watermark addr for overflow detection. */
> + void *stack_wmark_ofl;
Generally we try to avoid abbreviations unless really necessary.
Why not simply stack_overflow_watermark?
> /** Coro stack size. */
> size_t stack_size;
> /** Valgrind stack id. */
--
Konstantin Osipov, Moscow, Russia, +7 903 626 22 32
http://tarantool.io - www.twitter.com/kostja_osipov
More information about the Tarantool-patches
mailing list