summaryrefslogtreecommitdiff
path: root/service_solahart_jakarta_selatan__082122541663/system_crash_pflocal_fifo.mdwn
blob: 1dddc44ece536a789849855d0411341a259de090 (plain)
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
40
41
[[!meta copyright="Copyright © 2010 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_gnumach]]

IRC, unknown channel, unknown date:

`cat < /dev/zero | cat > /dev/null` will eventually make the system crash,
likewise when using a FIFO.

    <antrik> hm... VM activity seems much higher when running fifo than pfinet... may be the cause
    <antrik> "zero filled" and "page faults" are serveral times higher with pipe than with pfinet
    <antrik> (cow faults however are about the same...)
    <antrik> pflocal is about the same as fifo
    
    <antrik> no, because it usually takes like 20 minutes until it crashes, sometimes much longer
    
    <antrik> not sure, but the longest so far was in the range of hours IIRC
    
    <antrik> I think I never tested what happens on "cat /dev/zero >/dev/null"... another thing yet to try
    
    <antrik> Linux BTW seems to employ some major VM trickery in this case -- dd shows a transfer rate of 10 GB/s...
    
    <antrik> no, no anomalies in vmstat
    <antrik> the only observation I made is that number of page faults and some other number rise pretty quickly with pflocal and fifo, but not with pfinet
    <antrik> I guess that's somehow related to the fact that pfinet doesn't crash -- though I guess the difference is simply that pfinet is way slower...
    <antrik> (haven't checked that, though)
    
    <antrik> BTW, I'm not sure you got it right: the test case is "cat /dev/zero|cat >/dev/null", *not* "cat /dev/zero >/dev/null"
    
    <antrik> OK, "cat /dev/zero|tail -c 1" also crashes, so it's definitely not related to /dev/null
    <antrik> "dd if=/dev/zero|tail -c 1" crashes as well
    <antrik> but "tail -c 1 /dev/zero" doesn't seem to
    <antrik> cool... running multiple instances of the pipe test also considerably speeds up the crash