https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
Conrad Meyer changed:
What|Removed |Added
Blocks||232940
Referenced Bugs:
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
Conrad Meyer changed:
What|Removed |Added
Status|In Progress |Closed
Resolution|---
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
--- Comment #16 from Conrad Meyer ---
(In reply to Thomas Hurst from comment #15)
Thanks, that helps. It still seems like a weird way to go about it (smells
like a hand-rolled backup solution or something) but it's more concrete.
--
You
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
--- Comment #15 from Thomas Hurst ---
As for the use-case, here's the initial report:
> I'm seeing really, *REALLY* bad performance from stat(1) on FreeNAS. Case in
> point: a `find . -print0 | xargs -0 shasum >file.sha` takes about 7.5h
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
--- Comment #14 from Thomas Hurst ---
11.2-STABLE, consistent across several runs:
Conrad's patch: 8.792 real, 4.506 user, 4.788 sys; page: 0 hard/13520 soft,
swap: 0, I/O: 8818/0
My patch: 8.577 real, 4.547 user, 4.525 sys; page: 0
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
Conrad Meyer changed:
What|Removed |Added
Status|Open|In Progress
--- Comment #13 from
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
--- Comment #12 from commit-h...@freebsd.org ---
A commit references this bug:
Author: cem
Date: Sat Aug 11 02:56:44 UTC 2018
New revision: 337600
URL: https://svnweb.freebsd.org/changeset/base/337600
Log:
stat(1): cache id->name
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
--- Comment #11 from Conrad Meyer ---
Here's my non-scientific results for my proposed patch (on CURRENT GENERIC,
i.e., WITNESS and INVARIANTS are on, which introduces wider error bars than
without, but still):
BEFORE
conrad@n /u/s/u/stat
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
Conrad Meyer changed:
What|Removed |Added
Attachment #196041|0 |1
is obsolete|
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
--- Comment #9 from Conrad Meyer ---
And, heh, the need to cache id->name resolution was recognized *long* ago:
HISTORY
The user_from_uid() and group_from_gid() functions first appeared in
4.4BSD.
:-D.
--
You are receiving
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
--- Comment #8 from Conrad Meyer ---
(In reply to Garance A Drosehn from comment #5)
(In reply to Garance A Drosehn from comment #7)
> but why would using nscd be so much slower than doing a totally non-cached
> look
> up of the user and
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
--- Comment #7 from Garance A Drosehn ---
(In reply to Conrad Meyer from comment #4)
Just how much overhead is there in IPC? It's easy to see that a local cache
will beat the caching done by a separate nscd process, but why would using
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
--- Comment #6 from Garance A Drosehn ---
(In reply to Garance A Drosehn from comment #5)
for instance, on one sample directory:
time ls -ln
real0m0.010s
user0m0.001s
sys 0m0.007s
time ls -l# on host with nscd
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
--- Comment #5 from Garance A Drosehn ---
(In reply to Garance A Drosehn from comment #3)
Ugh. I didn't notice that 'ncsd' was mentioned in the response right before
mine! The performance of nscd might depend on how it's configured in
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
--- Comment #4 from Conrad Meyer ---
(In reply to Garance A Drosehn from comment #3)
See comment #2 :-). It's understandable given IPC overhead vs accessing a
static variable.
--
You are receiving this mail because:
You are the assignee
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
Garance A Drosehn changed:
What|Removed |Added
CC||g...@freebsd.org
--- Comment
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
Mark Linimon changed:
What|Removed |Added
Keywords||patch
--
You are receiving this
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
--- Comment #2 from Thomas Hurst ---
The built-in caching doesn't seem useful for this case. Enabling it results in
a test run creating this unfortunate entry in top:
63957 root 9 520 24000K 4696K uwait 19 9:15
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
Conrad Meyer changed:
What|Removed |Added
CC||c...@freebsd.org
--- Comment #1
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230491
Kubilay Kocak changed:
What|Removed |Added
Summary|[patch] stat(1): Improve|stat(1): Improve
20 matches
Mail list logo
