sysctl vs /proc/sys

sysctl vs /proc/sys

[YOSHIFUJI Hideaki / 吉藤英明]

Hello.

sysctl(2) seems to be deprecated, but
IMHO, /proc/sys cannot be the alternative
because it cannot return "previous" value atomicly.

I would say it is okay to leave sysctl(2)
because the interface is very common.
Or, I'd rather propose introducing sysctlbyname(2).

.. or do I miss something?

--yoshfuji

Re: sysctl vs /proc/sys

[James Morris]

On Fri, 12 Dec 2003, YOSHIFUJI Hideaki / [iso-2022-jp] ^[$B5HF#1QL@^[(B wrote:

> Or, I'd rather propose introducing sysctlbyname(2).

See the thread at 
http://marc.theaimsgroup.com/?l=linux-kernel&m=105788786819578&w=2



- James
-- 
James Morris
<jmorris@redhat.com>

Re: sysctl vs /proc/sys

[YOSHIFUJI Hideaki / 吉藤英明]

In article <Xine.LNX.4.44.0312120815170.2611-100000@thoron.boston.redhat.com> (at Fri, 12 Dec 2003 08:20:41 -0500 (EST)), James Morris <jmorris@redhat.com> says:

> On Fri, 12 Dec 2003, YOSHIFUJI Hideaki / [iso-2022-jp] ^[$B5HF#1QL@^[(B wrote:
> 
> > Or, I'd rather propose introducing sysctlbyname(2).
> 
> See the thread at 
> http://marc.theaimsgroup.com/?l=linux-kernel&m=105788786819578&w=2

Thanks, James. I think I missed the thread.

Well, some applications which can run on *BSD* and Linux uses sysctl.
Because using procfs changes the semantics, I'd love to have sysctlbyname().

Can we get the exact previous value from /proc/sys atomicly?
If so, I'm ok not to have sysctlbyname(2) 
because we can have sysctlbyname(3).

If no, please let's have sysctlbyname(2); there ARE
something we cannot do only without sysctl variants.

Note: sysctlbyname(3) and sysctlnametomib(3) is in FreeBSD.

--yoshfuji



 

Re: sysctl vs /proc/sys

[James Morris]

On Fri, 12 Dec 2003, YOSHIFUJI Hideaki / [iso-2022-jp] ^[$B5HF#1QL@^[(B wrote:

> Can we get the exact previous value from /proc/sys atomicly?

I'm not sure what you mean by this.


- James
-- 
James Morris
<jmorris@redhat.com>

Re: sysctl vs /proc/sys

[Mike Fedyk]

On Fri, Dec 12, 2003 at 09:28:32AM -0500, James Morris wrote:
> On Fri, 12 Dec 2003, YOSHIFUJI Hideaki / [iso-2022-jp] ^[$B5HF#1QL@^[(B wrote:
> 
> > Can we get the exact previous value from /proc/sys atomicly?
> 
> I'm not sure what you mean by this.

Read value and modify in one op?

Re: sysctl vs /proc/sys

[YOSHIFUJI Hideaki / 吉藤英明]

In article <20031212181517.GM15401@matchmail.com> (at Fri, 12 Dec 2003 10:15:17 -0800), Mike Fedyk <mfedyk@matchmail.com> says:

> On Fri, Dec 12, 2003 at 09:28:32AM -0500, James Morris wrote:
> > On Fri, 12 Dec 2003, YOSHIFUJI Hideaki / [iso-2022-jp] 吉藤英明 wrote:
> > 
> > > Can we get the exact previous value from /proc/sys atomicly?
> > 
> > I'm not sure what you mean by this.
> 
> Read value and modify in one op?

Yes, I meant to read old value and to set new one in one system call.
The sysctl system-call does lock_kernel, so sysctl(2) sytem-call 
are serialized.

How about procfs? Are there locks between multiple read-and-write?
Since It seems user-space cannot (or even should not) let kernel 
hold lock(s) during read-and-write operation, procfs cannot be 
the alternative.

Do I mis-understand something?

--yoshfuji

Re: sysctl vs /proc/sys

[H. Peter Anvin]

Followup to:  <20031213.094210.107050343.yoshfuji@linux-ipv6.org>
By author:    YOSHIFUJI Hideaki / =?iso-2022-jp?B?GyRCNUhGIzFRTEAbKEI=?= <yoshfuji@linux-ipv6.org>
In newsgroup: linux.dev.kernel
> 
> Yes, I meant to read old value and to set new one in one system call.
> The sysctl system-call does lock_kernel, so sysctl(2) sytem-call 
> are serialized.
> 

Dumb question... what do you need this for?  sysctl is mostly used to
initialization-time stuff...  In most cases you can use flock(),
probably, although that is advisory and not compulsory.

	-hpa
-- 
<hpa@transmeta.com> at work, <hpa@zytor.com> in private!
If you send me mail in HTML format I will assume it's spam.
"Unix gives you enough rope to shoot yourself in the foot."
Architectures needed: ia64 m68k mips64 ppc ppc64 s390 s390x sh v850 x86-64

Linux Kernel 2.6.0-test11 VFS problem

[Ananda Bhattacharya]

	Hi, I am trying to get kernel 2.6.0-test11 on my 
laptop. this is really stupid, I got all the "latest" 
utilities etc... etc... etc... 

But when I boot i get this error


###################################
VFS Cannot open root device "hda5" 
Please append a correct "root=" boot option 
Kernel Panic: VFS: Unable to mount root filesystem on hda5 
###################################



here is my grub.conf file

*******************************
title 2.6-test-11
        root (hd0,1)
        kernel /vmlinuz-2.6-test-11 ro root=305
********************************


What am I doing wrong today ??


I know this is something really simple somwhere, but I am 
giving this up and hitting up and hitting a Pilsner 
Urquell.....

 -- 
Recruiter: "How do you write a perl script?"
Canidate : "You type it."
	-Actual phone interview, Circa Nov 2003

Re: Linux Kernel 2.6.0-test11 VFS problem

[Mike Fedyk]

On Fri, Dec 12, 2003 at 09:06:43PM -0500, Ananda Bhattacharya wrote:
> ###################################
> VFS Cannot open root device "hda5" 
> Please append a correct "root=" boot option 
> Kernel Panic: VFS: Unable to mount root filesystem on hda5 
> ###################################

What filesystem is on hda5? and did you specify it as /dev/hda5? and did you  
compile that filesystem into your kernel (ie, not as a module)?