[BUG] 5.0.0_beta1: does not auto-unmount

[BUG] 5.0.0_beta1: does not auto-unmount

[Denys Vlasenko]

[-- Attachment #1: Type: text/plain, Size: 1658 bytes --]

Hi Ian,

I run automount like this:

# automount -f -s -v --timeout 15 /.local/var/service/automount/master.file

My master.file is one-liner:

/mnt/auto program:/.local/var/service/automount/mapper.sh

mapper.sh is attached.

This setup (with this exact mapper.sh) worked flawlessly
with 4.1.4, but with 5.0.0_beta1 I have a problem:

# automount -f -s -v --timeout 15 /.local/var/service/automount/master
Starting automounter version 5.0.0_beta1, master map /.local/var/service/automount/master
using kernel protocol version 5.00
using timeout 15 seconds; freq 4 secs

I do "cd /mnt/auto/vfat.sdb1" (sdb1 is my USB memory stick)
and automount says:

attempting to mount entry /.local/mnt/auto/vfat.sdb1
mounted /mnt/auto/vfat.sdb1

So far so good, mount shows:
/dev/sdb1 on /mnt/auto/vfat.sdb1 type vfat (rw,nosuid,noexec,fmask=0000,dmask=0000,codepage=cp866,iocharset=koi8-r,shortname=win95,quiet,uni_xlate)

But it never auto-unmounts. I am 100% sure nothing
is keeping files or directories busy on /dev/sdb1 (I cd'ed away
from the directory). Automount is not saying anything,
strace -p <PID> shows one line:

rt_sigtimedwait(~[CHLD CONT RTMIN RT_1],

I killed 4 hours of my time making automount 5.0.0 even start on my machine,
and it's 4:00 am here, so at this time I am a rather unhappy panda and
probably I'll go back to 4.1.4 instead of chasing down yet another bug.

# size autofs*/sbin/*
   text    data     bss     dec     hex filename
  35143     916   10084   46143    b43f autofs-4.1.4_beta1/sbin/automount
 118494    1344     884  120722   1d792 autofs-5.0.0_beta1/sbin/automount

During 4.1.4 -> 5.0.0 size has *tripled*. Uncool. :(
--
vda

[-- Attachment #2: mapper.sh --]
[-- Type: application/x-shellscript, Size: 1607 bytes --]

[-- Attachment #3: Type: text/plain, Size: 140 bytes --]

_______________________________________________
autofs mailing list
autofs@linux.kernel.org
http://linux.kernel.org/mailman/listinfo/autofs

Re: [BUG] 5.0.0_beta1: does not auto-unmount

[Ian Kent]

On Wed, 2007-09-26 at 11:23 +0100, Denys Vlasenko wrote:
> Hi Ian,
> 
> I run automount like this:
> 
> # automount -f -s -v --timeout 15 /.local/var/service/automount/master.file
> 
> My master.file is one-liner:
> 
> /mnt/auto program:/.local/var/service/automount/mapper.sh
> 
> mapper.sh is attached.
> 
> This setup (with this exact mapper.sh) worked flawlessly
> with 4.1.4, but with 5.0.0_beta1 I have a problem:
> 
> # automount -f -s -v --timeout 15 /.local/var/service/automount/master
> Starting automounter version 5.0.0_beta1, master map /.local/var/service/automount/master
> using kernel protocol version 5.00
> using timeout 15 seconds; freq 4 secs
> 
> I do "cd /mnt/auto/vfat.sdb1" (sdb1 is my USB memory stick)
> and automount says:
> 
> attempting to mount entry /.local/mnt/auto/vfat.sdb1
> mounted /mnt/auto/vfat.sdb1
> 
> So far so good, mount shows:
> /dev/sdb1 on /mnt/auto/vfat.sdb1 type vfat (rw,nosuid,noexec,fmask=0000,dmask=0000,codepage=cp866,iocharset=koi8-r,shortname=win95,quiet,uni_xlate)
> 
> But it never auto-unmounts. I am 100% sure nothing
> is keeping files or directories busy on /dev/sdb1 (I cd'ed away
> from the directory). Automount is not saying anything,
> strace -p <PID> shows one line:

Strange, I'll try and duplicate this and see how far I get.
If you do want to make sure nothing is asking it to remount enable debug
and ensure facility daemon.* is being logged. If there are accesses
keeping it open the requesting pid will be logged.

> 
> rt_sigtimedwait(~[CHLD CONT RTMIN RT_1],
> 
> I killed 4 hours of my time making automount 5.0.0 even start on my machine,
> and it's 4:00 am here, so at this time I am a rather unhappy panda and
> probably I'll go back to 4.1.4 instead of chasing down yet another bug.

Ya but I'm getting there.
The change is huge to be sure, and many things do work well.

> 
> # size autofs*/sbin/*
>    text    data     bss     dec     hex filename
>   35143     916   10084   46143    b43f autofs-4.1.4_beta1/sbin/automount
>  118494    1344     884  120722   1d792 autofs-5.0.0_beta1/sbin/automount
> 
> During 4.1.4 -> 5.0.0 size has *tripled*. Uncool. :(

Not sure what I can do about that, sorry.

Re: [BUG] 5.0.0_beta1: does not auto-unmount

[Ian Kent]

Denys,

On another note.
5.0.0_beta1?

Since beta1 there were five other beta and three rc releases. The
initial release was actually 5.0.1 and were at 5.0.2 now with a bunch of
corrections which need to be consolidated into 5.0.3 time soon. There
have been many corrections including ones for expire issues.

Ian

Re: [BUG] 5.0.0_beta1: does not auto-unmount

[Denys Vlasenko]

On Wednesday 26 September 2007 16:34, Ian Kent wrote:
> 
> Denys,
> 
> On another note.
> 5.0.0_beta1?

It's the latest thing I found at 

http://www.kernel.org/pub/linux/daemons/autofs/v4/

Aha. I see. I had to fetch stuff from

http://www.kernel.org/pub/linux/daemons/autofs/v5/

right?

--
vda

Re: [BUG] 5.0.0_beta1: does not auto-unmount

[Ian Kent]

On Wed, 2007-09-26 at 17:00 +0100, Denys Vlasenko wrote:
> On Wednesday 26 September 2007 16:34, Ian Kent wrote:
> > 
> > Denys,
> > 
> > On another note.
> > 5.0.0_beta1?
> 
> It's the latest thing I found at 
> 
> http://www.kernel.org/pub/linux/daemons/autofs/v4/
> 
> Aha. I see. I had to fetch stuff from
> 
> http://www.kernel.org/pub/linux/daemons/autofs/v5/
> 
> right?

Yep, that's the place.
Ian

Re: [BUG] 5.0.0_beta1: does not auto-unmount

[Denys Vlasenko]

On Wednesday 26 September 2007 17:36, Ian Kent wrote:
> On Wed, 2007-09-26 at 17:00 +0100, Denys Vlasenko wrote:
> > On Wednesday 26 September 2007 16:34, Ian Kent wrote:
> > > 
> > > Denys,
> > > 
> > > On another note.
> > > 5.0.0_beta1?
> > 
> > It's the latest thing I found at 
> > 
> > http://www.kernel.org/pub/linux/daemons/autofs/v4/
> > 
> > Aha. I see. I had to fetch stuff from
> > 
> > http://www.kernel.org/pub/linux/daemons/autofs/v5/
> > 
> > right?
> 
> Yep, that's the place.
> Ian

autofs-5.0.2 still references www.kernel.org/pub/linux/daemons/autofs/v4/:

# grep -r /v4 .
./INSTALL:     from http://www.kernel.org/pub/linux/daemons/autofs/v4.
./autofs.spec:Source: ftp://ftp.kernel.org/pub/linux/daemons/autofs/v4/autofs-%{version}.tar.gz
./gentoo/net-fs/autofs/autofs-5.0.2.ebuild:SRC_URI_BASE="mirror://kernel/linux/daemons/${PN}/v4"

What do you think about removing autofs-5.0.0_beta1.tar.* from .../v4 directory
in order to prevent confusion?
--
vda

Re: [BUG] 5.0.0_beta1: does not auto-unmount

[Ian Kent]

On Thu, 2007-09-27 at 21:14 +0100, Denys Vlasenko wrote:
> On Wednesday 26 September 2007 17:36, Ian Kent wrote:
> > On Wed, 2007-09-26 at 17:00 +0100, Denys Vlasenko wrote:
> > > On Wednesday 26 September 2007 16:34, Ian Kent wrote:
> > > > 
> > > > Denys,
> > > > 
> > > > On another note.
> > > > 5.0.0_beta1?
> > > 
> > > It's the latest thing I found at 
> > > 
> > > http://www.kernel.org/pub/linux/daemons/autofs/v4/
> > > 
> > > Aha. I see. I had to fetch stuff from
> > > 
> > > http://www.kernel.org/pub/linux/daemons/autofs/v5/
> > > 
> > > right?
> > 
> > Yep, that's the place.
> > Ian
> 
> autofs-5.0.2 still references www.kernel.org/pub/linux/daemons/autofs/v4/:
> 
> # grep -r /v4 .
> ./INSTALL:     from http://www.kernel.org/pub/linux/daemons/autofs/v4.
> ./autofs.spec:Source: ftp://ftp.kernel.org/pub/linux/daemons/autofs/v4/autofs-%{version}.tar.gz
> ./gentoo/net-fs/autofs/autofs-5.0.2.ebuild:SRC_URI_BASE="mirror://kernel/linux/daemons/${PN}/v4"

Ha, missed that completely.
Thnks.

> 
> What do you think about removing autofs-5.0.0_beta1.tar.* from .../v4 directory
> in order to prevent confusion?

Yes, I'd like to but I was a bit concerned that distros that pull these
tarballs might expect them to be present.

Maybe I could encourage an update if I did, which would be desirable.

Ian

Re: [BUG] 5.0.0_beta1: does not auto-unmount

[Denys Vlasenko]

> > autofs-5.0.2 still references www.kernel.org/pub/linux/daemons/autofs/v4/:
> > 
> > # grep -r /v4 .
> > ./INSTALL:     from http://www.kernel.org/pub/linux/daemons/autofs/v4.
> > ./autofs.spec:Source: ftp://ftp.kernel.org/pub/linux/daemons/autofs/v4/autofs-%{version}.tar.gz
> > ./gentoo/net-fs/autofs/autofs-5.0.2.ebuild:SRC_URI_BASE="mirror://kernel/linux/daemons/${PN}/v4"
> 
> Ha, missed that completely.
> Thnks.
> 
> > 
> > What do you think about removing autofs-5.0.0_beta1.tar.* from .../v4 directory
> > in order to prevent confusion?
> 
> Yes, I'd like to but I was a bit concerned that distros that pull these
> tarballs might expect them to be present.
> 
> Maybe I could encourage an update if I did, which would be desirable.

I mean, by removing 5.0.0beta from v4 directory you will avoid having
confused users thinking that 5.0.0_beta1 is latest.

I tried 5.0.2 today, summary:

-s --stderr       missing
-f --foreground   works, not mentioned in help text

I didn't go further, can't say whether it works for me.
Will test 5.0.3 whenever you want.
--
vda

Re: [BUG] 5.0.0_beta1: does not auto-unmount

[Ian Kent]

On Sun, 2007-09-30 at 02:25 +0100, Denys Vlasenko wrote:
> > > autofs-5.0.2 still references www.kernel.org/pub/linux/daemons/autofs/v4/:
> > > 
> > > # grep -r /v4 .
> > > ./INSTALL:     from http://www.kernel.org/pub/linux/daemons/autofs/v4.
> > > ./autofs.spec:Source: ftp://ftp.kernel.org/pub/linux/daemons/autofs/v4/autofs-%{version}.tar.gz
> > > ./gentoo/net-fs/autofs/autofs-5.0.2.ebuild:SRC_URI_BASE="mirror://kernel/linux/daemons/${PN}/v4"
> > 
> > Ha, missed that completely.
> > Thnks.
> > 
> > > 
> > > What do you think about removing autofs-5.0.0_beta1.tar.* from .../v4 directory
> > > in order to prevent confusion?
> > 
> > Yes, I'd like to but I was a bit concerned that distros that pull these
> > tarballs might expect them to be present.
> > 
> > Maybe I could encourage an update if I did, which would be desirable.
> 
> I mean, by removing 5.0.0beta from v4 directory you will avoid having
> confused users thinking that 5.0.0_beta1 is latest.

Yes, I think I will remove all the v5 specific stuff from the v4
directory at the same time I post 5.0.3.

It's been quite a while and people that have version 5 should be aware
of updates and really should be using later versions.

> 
> I tried 5.0.2 today, summary:
> 
> -s --stderr       missing
> -f --foreground   works, not mentioned in help text

Are you sure you applied the additional foreground logging patch (well
actually if you test 5.0.2 you should really apply all the patches
listed in patch_order-5.0.2 to get a "current state" view).

Are you the same person that originally contributed these patches?
I have the name "Denis Vlasenko" in the original patch emails.

Anyway, I never quite completely apply the patches but what we have now
is fairly close. In particular I haven't provided a distinct "log to
stderr". Over the considerable time this has been present I've rarely
needed to use the foreground option but when I have I've always wanted
to send the log to my tty. Especially since you can just redirect the
output if you need to when running in foreground.

Ian

Re: [BUG] 5.0.0_beta1: does not auto-unmount

[Denys Vlasenko]

On Tuesday 02 October 2007 07:09, Ian Kent wrote:
> > I tried 5.0.2 today, summary:
> > 
> > -s --stderr       missing
> > -f --foreground   works, not mentioned in help text
> 
> Are you sure you applied the additional foreground logging patch (well
> actually if you test 5.0.2 you should really apply all the patches
> listed in patch_order-5.0.2 to get a "current state" view).

Hmmm... no.

> Are you the same person that originally contributed these patches?
> I have the name "Denis Vlasenko" in the original patch emails.

Yes. "Denys" is my country's officialdom's idea how to transliterate
my name into Latin letters. [idiots]

> Anyway, I never quite completely apply the patches but what we have now
> is fairly close. In particular I haven't provided a distinct "log to
> stderr". Over the considerable time this has been present I've rarely
> needed to use the foreground option but when I have I've always wanted
> to send the log to my tty. Especially since you can just redirect the
> output if you need to when running in foreground.

I don't understand you. Currently, even it --foreground mode
automount logs everything to syslog, IIRC.
Hot do you redirect that? To where?

I want to be able to direct log to be on stderr or stdout
(I log each daemon's output separately by capturing those).
--
vda

Re: [BUG] 5.0.0_beta1: does not auto-unmount

[Ian Kent]

On Tue, 2007-10-02 at 10:33 +0100, Denys Vlasenko wrote:
> On Tuesday 02 October 2007 07:09, Ian Kent wrote:
> > > I tried 5.0.2 today, summary:
> > > 
> > > -s --stderr       missing
> > > -f --foreground   works, not mentioned in help text
> > 
> > Are you sure you applied the additional foreground logging patch (well
> > actually if you test 5.0.2 you should really apply all the patches
> > listed in patch_order-5.0.2 to get a "current state" view).
> 
> Hmmm... no.
> 
> > Are you the same person that originally contributed these patches?
> > I have the name "Denis Vlasenko" in the original patch emails.
> 
> Yes. "Denys" is my country's officialdom's idea how to transliterate
> my name into Latin letters. [idiots]

HaHaha, right.

> 
> > Anyway, I never quite completely apply the patches but what we have now
> > is fairly close. In particular I haven't provided a distinct "log to
> > stderr". Over the considerable time this has been present I've rarely
> > needed to use the foreground option but when I have I've always wanted
> > to send the log to my tty. Especially since you can just redirect the
> > output if you need to when running in foreground.
> 
> I don't understand you. Currently, even it --foreground mode
> automount logs everything to syslog, IIRC.
> Hot do you redirect that? To where?

The recent patch for foreground logging will send everything to stderr
unconditionally. I should have realized you wouldn't get this bit,
sorry.

> 
> I want to be able to direct log to be on stderr or stdout
> (I log each daemon's output separately by capturing those).

Yep, got it.
Hopefully the patch I mentioned above will do what's needed.

Ian