Hi,
while watching bcache sysfs values, I noticed that sometimes the
/sys/fs/bcache/<cset-uuid>/congested
shows non-zero, indicating that my SSD cannot cope with all the
writes.
I tried to understand what this value means (not mentioned in
Documentation/bcache.txt).
Using kernel 3.13-rc3.
In sysfs.c:
510 sysfs_hprint(congested,
511 ((uint64_t) bch_get_congested(c)) << 9);
In request.c:
464 /* Congested? */
465
466 unsigned bch_get_congested(struct cache_set *c)
467 {
468 int i;
469 long rand;
470
471 if (!c->congested_read_threshold_us &&
472 !c->congested_write_threshold_us)
473 return 0;
474
475 i = (local_clock_us() - c->congested_last_us) / 1024;
476 if (i < 0)
477 return 0;
478
479 i += atomic_read(&c->congested);
480 if (i >= 0)
481 return 0;
482
483 i += CONGESTED_MAX;
484
485 if (i > 0)
486 i = fract_exp_two(i, 6);
487
488 rand = get_random_int();
489 i -= bitmap_weight(&rand, BITS_PER_LONG);
490
491 return i > 0 ? i : 1;
492 }
Apparently c->congested_last_us is a value in microseconds, i and
c->congested are values in (approx.) milliseconds.
But after line 485 I am completely lost :-)
Line 485 is something like
i = int (2 ** (i/64.0));
And then in line 489 a random number <= 64 (or <= 32 on 32 bit) is
subtracted.
What does it mean? What unit is "congested" in?
Regards
Matthias
--
To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
