1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
|
[[!meta copyright="Copyright © 2011 Free Software Foundation, Inc."]]
[[!meta license="""[[!toggle id="license" text="GFDL 1.2+"]][[!toggleable
id="license" text="Permission is granted to copy, distribute and/or modify this
document under the terms of the GNU Free Documentation License, Version 1.2 or
any later version published by the Free Software Foundation; with no Invariant
Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license
is included in the section entitled [[GNU Free Documentation
License|/fdl]]."]]"""]]
[[!tag open_issue_hurd]]
From: Samuel Thibault <samuel.thibault@gnu.org>
Subject: rm -fr slowness
I have always been surprised by the slowness of a mere rm -fr. Looking a
bit inside, I see that diskfs_dirremove_hard() calls diskfs_file_update
(dp, 1) (as does diskfs_truncate, diskfs_direnter_hard, and
diskfs_dirrewrite_hard). diskfs_file_update then calls pager_sync on
the pager, which thus writes back the whole ext2fs pager!
This sounds a bit excessive to me, an unlink could just record it in
memory and actually sync later. Also, the wait flag is set, so we
really waits for all I/Os, which basically means strictly serializing
file removals: remove one file, wait for the disk to have done it
(~10ms), remove the next one, etc. I guess this is for safety reasons
against crashes, but isn't the sync option there for such kind of
# IRC, freenode, #hurd, 2011-07-23
<antrik> youpi: hm... async deletion does have one downside: I just removed
something to make space, and retried the other command immediately
afterwards, and it still said "no space left on device"... a few seconds
later (after the next regular sync I suppose?) it worked
<youpi> well, that's sorta expected, yes
<youpi> we get the same on Linux
<youpi> Mmm, on second thought, I'm not sure how that can happen
<youpi> the asynchronous thing is for disk writes, not cache writes
|
