Hans Reiser <[email protected]> wrote:
>
> I suspect that when someone did the search and replace when creating
> balance_dirty_pages_ratelimited_nr they failed to read the code and
> realize this code path was already effectively ratelimited. The result
> is they made it excessively infrequent (every 1MB if ratelimit is 8) in
> its calling balance_dirty_pages.
?? There's been no change to balance_dirty_pages_ratelimited(). I merely
widened the interface a bit: introduced the new
balance_dirty_pages_ratelimited_nr() and did
static inline void
balance_dirty_pages_ratelimited(struct address_space *mapping)
{
balance_dirty_pages_ratelimited_nr(mapping, 1);
}
That being said, if reiserfs has `number of pages' in hand then yes, it
makes sense to migrate over to balance_dirty_pages_ratelimited_nr().
> If anyone has ever seen this as an actual problem on a real system, I
> would be curious to hear of it.
No, I wouldn't expect it to make much difference.
All that gunk is there just to avoid calling the expensive
get_writeback_state() once per set_page_dirty().
Inaccuracy here will introduce the possibility that we'll transiently dirty
more memory than dirty_ratio permits, but it'll only be a little bit (eight
times the amount of memory which is dirtied per balance_dirty_pages_ratelimited()
call).
That's a small amount of memory. But if you have 1000 filesystems mounted
and they all do the same thing at the same time, things could get a bit
sticky. Your patch will (greatly) reduce the possibility of even that
scenario.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[Index of Archives]
[Kernel Newbies]
[Netfilter]
[Bugtraq]
[Photo]
[Stuff]
[Gimp]
[Yosemite News]
[MIPS Linux]
[ARM Linux]
[Linux Security]
[Linux RAID]
[Video 4 Linux]
[Linux for the blind]
[Linux Resources]
