Nabble - Fuse - Dev

Addition to the list FileSystems: RevealFS

2009-12-15T15:54:56Z

Hello!

Please add RevealFS (details below) to the list of file systems on [1]
for me as I have no write permissions to it. From browsing the whole
tree today I think RevealFS does not fit into any of the categories.
RevealFS is most similar to rofs-filtered which is listed on the main
page, too.

Author:
Sebastian Pipping / sebastian at pipping org

Homepage:
http://git.freitagsrunde.org/?p=revealfs.git;a=summary

Description:
Bind-mounts a folder in read-only mode and hides all files and
folders that do not have the extended file system attribute (xattr)
"user.public" set. Can be used to reveal a subset of files to the
public while keeping both internal and public documents together in
the same tree. Written in Python. See also: rofs-filtered.

Thanks in advance!

Best regards,

Sebastian

[1] https://sourceforge.net/apps/mediawiki/fuse/index.php?title=FileSystems

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

small bug in fuse-python 0.2

2009-12-15T15:52:32Z

hello fuse-devel list!

the latest release of fuse-python (version 0.2) has a bug with the
constructors of the classes Flock and Timespec: Calls like

Timespec(tv_sec = acc_sec, tv_nsec = acc_nsec)

are made against these constructors:

class Flock(FuseStruct):
def __init__(self, name, **kw):
[..snip..]
FuseStruct.__init__(self, **kw)

class Timespec(FuseStruct):
def __init__(self, name, **kw):
[..snip..]
FuseStruct.__init__(self, **kw)

someone fixed that in CVS with a patch like

- def __init__(self, name, **kw):
+ def __init__(self, **kw):

while that makes the internal calls from fuse-python work it would break
any previously valid external calls like

Timespec('foobar', tv_sec = acc_sec, tv_nsec = acc_nsec)

depending on whether you want name to go to FuseStruct.__init__ i would
like to request repatching to either

+ def __init__(self, name=None, **kw):

or

+ def __init__(self, name=None, **kw):
+ kw['name'] = name

i hope that makes the problem clear.

==================================================
please:

- let me know which variant you prefer
(so i can fix it it downstream at gentoo linux)
- repatch upstream
- make a new release of fuse-python after

thank you!
==================================================

a new release is needed as code out there is really hitting the bug;
i ran into this with the pytagsfs test suite:

Traceback (most recent call last):
File "/usr/lib/python2.6/site-packages/fuse.py",\
line 361, in __call__
return apply(self.func, args, kw)
File "/usr/lib/python2.6/site-packages/fuse.py",\
line 782, in wrap\
ts_acc = Timespec(tv_sec = acc_sec, tv_nsec = acc_nsec)
TypeError: __init__() takes exactly 2 non-keyword arguments (1 given)

looking forward to your feedback!

sebastian

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Big write requests

2009-12-13T01:10:40Z

Hi,

I've just tested the 2.8.1. And would like to know in which conditions big writes are requested ?
Must I configure FUSE to privilege the big writes ?
Or would it be a kernel policy that I can't handle ?

Thanks all
Regards
Didier
------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: fuse_entry_param

2009-12-12T18:43:23Z

On 2009-12-12, at 8:04 PM, Nikolaus Rath wrote:

>> This is used by NFS: the NFS clients will refer to files by file
>> handle (a value made up of the inode number and generation number). If
>> the generation number changes (without the knowledge of the NFS
>> clients), then the NFS client will receive ESTALE if it tries to
>> access the old file so that it doesn't break the new file.
>
> I see. Do you know how other file systems handle the generation number
> in practice?

No, but I would imagine that it's like having a larger inode number (ie: if the file-system stores a 32-bit inode number, then the generation number can be the 'upper 32-bits' of a 64-bit NFS file handle that is made up of the inode number and generation number). So it is really optional. For example, my file-system has 64-bit inode numbers (I modified fuse_ino_t so that it is 64-bit - I can get away with this since it is implemented in an embedded system), and it doesn't reuse inode numbers, so the generation number returned is always 1.

> I guess as long as the file system has a static inode
> table like ext3, one can store the generation no for each inode in
> that table as well. But what if inodes are dynamically allocated? It
> seems to me that I would have to keep a table of every inode ever used
> to save the last generation numbers...

Lifetime refers to the existence of the file-system, and not the duration of the mount. Inode numbers should be stored in the file-system permanently and refer to only one object (there can be multiple directory entries (hard links) to the same inode).

>From and NFS perspective, the point is that the same files can be accessed over and over again regardless of whether the NFS server (your file-system) is restarted (unmounted and mounted, server crash, reboot for patching, etc.) while the NFS client has a handle into the file-system.

Just don't change the generation of the root directory in your file-system, because that would require that the NFS clients must unmount and mount again in order to access your file-system over an NFS server restart.

Regards,

John.

--
John Muir - john@...

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: fuse_entry_param

2009-12-12T17:04:27Z

John Muir <john@...> writes:

>> How do I have to use the "generation" attribute? The documentation
>> in fuse_lowlevel.h says that "The ino/generation pair should be
>> unique for the filesystem's lifetime". But what exactly does unique
>> mean here? If I have a file with inode 2, and then delete that file
>> and later I create a new file that occupies the same inode, do I
>> have to increase the generation?
>
>
> Yes.
>
> This is used by NFS: the NFS clients will refer to files by file
> handle (a value made up of the inode number and generation number). If
> the generation number changes (without the knowledge of the NFS
> clients), then the NFS client will receive ESTALE if it tries to
> access the old file so that it doesn't break the new file.

I see. Do you know how other file systems handle the generation number
in practice? I guess as long as the file system has a static inode
table like ext3, one can store the generation no for each inode in
that table as well. But what if inodes are dynamically allocated? It
seems to me that I would have to keep a table of every inode ever used
to save the last generation numbers...

Does "lifetime" refer to the duration of the mount or the existence of
the file system?

Best,

-Nikolaus

--
»Time flies like an arrow, fruit flies like a Banana.«

PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: fuse_entry_param

2009-12-12T12:30:32Z

> How do I have to use the "generation" attribute? The documentation in
> fuse_lowlevel.h says that "The ino/generation pair should be unique for
> the filesystem's lifetime". But what exactly does unique mean here? If I
> have a file with inode 2, and then delete that file and later I create a
> new file that occupies the same inode, do I have to increase the
> generation?

Yes.

This is used by NFS: the NFS clients will refer to files by file handle (a value made up of the inode number and generation number). If the generation number changes (without the knowledge of the NFS clients), then the NFS client will receive ESTALE if it tries to access the old file so that it doesn't break the new file.

Similarly, this applies the 'FSID' of the file-system itself with NFS, but unfortunately we don't control the FSID. We should change fuse to allow us to set the FSID from the file-system.

> Entries and attributes in my file systems are only changed in response
> to requests from the FUSE kernel module. Does that mean that I can
> effectively set attr_timeout and entry_timeout to infinity?

Yes. In my file-system, the time-out is set to 3600 seconds, although it could be infinite. Since the file-system itself can introduce changes, I use the fuse_lowlevel_notify_inval_inode/entry() functions to invalidate the attributes in the kernel.

Regards,

John.

--
John Muir - john@...

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: fuse_entry_param

2009-12-12T09:07:52Z

Hi,

Is there really no one able to answer this (and my other mail about the
arguments to fuse_mount and fuse_lowlevel_new) at the moment? I'm sure
there are some other people here who use the low-level API... or are
they all on their (deserved) holidays? *g*

Best,
Nikolaus

Nikolaus Rath <Nikolaus@...> writes:

> Hello,
>
> I have some questions about the fuse_entry_param structure:
>
> How do I have to use the "generation" attribute? The documentation in
> fuse_lowlevel.h says that "The ino/generation pair should be unique for
> the filesystem's lifetime". But what exactly does unique mean here? If I
> have a file with inode 2, and then delete that file and later I create a
> new file that occupies the same inode, do I have to increase the
> generation?
>
>
> Entries and attributes in my file systems are only changed in response
> to requests from the FUSE kernel module. Does that mean that I can
> effectively set attr_timeout and entry_timeout to infinity? If not, what
> value should I use?
>
>
> Thanks!
>
> -Nikolaus

-Nikolaus

--
»Time flies like an arrow, fruit flies like a Banana.«

PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: FW: Re: call of fuse_main() in a application

2009-12-11T12:17:42Z

Mylius WTW <Mylius_M@...> writes:

>> To answer your last question: your program *is* running normally the
>> whole time. If you want your program to do other things in parallel to
>> serving file system requests, you need to use threads or multiple
>> processes.
>
>
> I use already a lot of threads. Also fuse_main is started in a own thread.
> But nevertheless the whole application comes then daemonizes after launch
> of fuse in one of the threads from the application. For this reasen I guess I
> must take a own task (process) for fuse.

No, you don't have to. But daemonization affects the entire process, not
just the current thread. See e.g.
http://www.google.com/cse?q=unix+daemonize&sa=Search second link.

-Nikolaus

--
»Time flies like an arrow, fruit flies like a Banana.«

PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

FW: Re: call of fuse_main() in a application

2009-12-11T08:05:08Z

> To answer your last question: your program *is* running normally the
> whole time. If you want your program to do other things in parallel to
> serving file system requests, you need to use threads or multiple
> processes.

I use already a lot of threads. Also fuse_main is started in a own thread.
But nevertheless the whole application comes then daemonizes after launch
of fuse in one of the threads from the application. For this reasen I guess I
must take a own task (process) for fuse.

Maik

>
> > Mylius WTW <Mylius_M@...> writes:
> > > Hi Nikolaus,
> > >
> > >> No, fuse_main just daemonizes your program. It is still running, and
> > >> after you unmount the file system, fuse_main will return and program
> > >> execution will continue normally.
> > >
> > > Thanks, that is right. But how can I call 'fuse_main()' in my
> > > application without daemonizes of my program, so that my program runs
> > > the whole time normally?
> >
> > Don't get me wrong, but judging from the questions that you have been
> > asking, whatever you are trying to do at the moment is a bit too
> > ambitious for you right now.
> >
> > Why don't you try to get familiar with C programming first, *then* get
> > familiar with unix file systems, and *then* start writing your own file
> > system in FUSE?
> >
> >
> > (To answer your last question: your program *is* running normally the
> > whole time. If you want your program to do other things in parallel to
> > serving file system requests, you need to use threads or multiple
> > processes.)
> >
> >
> > Best,
> >
> > -Nikolaus
> >
> > --
> > »Time flies like an arrow, fruit flies like a Banana.«
> >
> > PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C
> >
> >
> > ------------------------------------------------------------------------------
> > Return on Information:
> > Google Enterprise Search pays you back
> > Get the facts.
> > http://p.sf.net/sfu/google-dev2dev
> > _______________________________________________
> > fuse-devel mailing list
> > fuse-devel@...
> > https://lists.sourceforge.net/lists/listinfo/fuse-devel
> >
>
>

______________________________________________________
GRATIS für alle WEB.DE-Nutzer: Die maxdome Movie-FLAT!
Jetzt freischalten unter http://movieflat.web.de

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: Know last read of sequence of reads

2009-12-11T05:50:24Z

Daniel García Coego <flamealchemist86@...> writes:
> Hi, I'm designing a file system I'll implement with fuse and I wonder if
> you can know what is the last read of a sequence, but not the last read
> of the file. For example, when you open a large PDF, several reads are
> called, but only the ones that are needed, not the whole file. I want to
> know if there's a way to recognize the last read of that sequence (hope
> it's clear). It would be very helpful for me.

It's not clear to me what 'sequence' you mean, but the answer to your
questions is almost certainly no, it is not possible. Just ask yourself
how you would implement anything like that.

Best,

-Nikolaus

--
»Time flies like an arrow, fruit flies like a Banana.«

PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: call of fuse_main() in a application

2009-12-11T05:47:43Z

Mylius WTW <Mylius_M@...> writes:
> Hi Nikolaus,
>
>> No, fuse_main just daemonizes your program. It is still running, and
>> after you unmount the file system, fuse_main will return and program
>> execution will continue normally.
>
> Thanks, that is right. But how can I call 'fuse_main()' in my
> application without daemonizes of my program, so that my program runs
> the whole time normally?

Don't get me wrong, but judging from the questions that you have been
asking, whatever you are trying to do at the moment is a bit too
ambitious for you right now.

Why don't you try to get familiar with C programming first, *then* get
familiar with unix file systems, and *then* start writing your own file
system in FUSE?

(To answer your last question: your program *is* running normally the
whole time. If you want your program to do other things in parallel to
serving file system requests, you need to use threads or multiple
processes.)

Best,

-Nikolaus

--
»Time flies like an arrow, fruit flies like a Banana.«

PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Know last read of sequence of reads

2009-12-11T04:47:15Z

Hi, I'm designing a file system I'll implement with fuse and I wonder if
you can know what is the last read of a sequence, but not the last read
of the file. For example, when you open a large PDF, several reads are
called, but only the ones that are needed, not the whole file. I want to
know if there's a way to recognize the last read of that sequence (hope
it's clear). It would be very helpful for me.

Thanks.

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: call of fuse_main() in a application

2009-12-11T03:44:06Z

Hello Maik,

> > No, fuse_main just daemonizes your program. It is still running,
> > and after you unmount the file system, fuse_main will return and
> > program execution will continue normally.
>
> Thanks, that is right. But how can I call 'fuse_main()' in my
> application without daemonizes of my program, so that my program
> runs the whole time normally?

Well, essentially you need to write your own version of fuse_main()
--- how to do this was discussed elsewhere on this list (under the
heading of threads or somesuch) and ought to be locate-able in the
archives.

Sven
--
_ ___ ___ ___
__| |/ __|| __|/ __| The dCache File System
/ _` | (__ | _| \__ \ An archive file-system for PB of data
\__,_|\___||_| |___/ http://www.desy.de/~utcke/Data/

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: call of fuse_main() in a application

2009-12-11T02:57:01Z

Hi Nikolaus,

> No, fuse_main just daemonizes your program. It is still running, and
> after you unmount the file system, fuse_main will return and program
> execution will continue normally.

Thanks, that is right. But how can I call 'fuse_main()' in my application
without daemonizes of my program, so that my program runs the whole time normally?

Maik

> -----Ursprüngliche Nachricht-----
> Von: "Nikolaus Rath" <Nikolaus@...>
> Gesendet: 09.12.09 20:15:11
> An: fuse-devel@...
> Betreff: Re: [fuse-devel] call of fuse_main() in a application

> Mylius WTW <Mylius_M@...> writes:
> > Hallo,
> >
> > I want to call the function in a application where
> > a lot of other threads are also running. If I call the
> > fuse_main function with static parameters like:
> >
> > func_xyz()
> > {
> > ...
> > int argc = 2;
> > char *argv[] =
> > {
> > "./fs_net",
> > "mountdir"
> > };
> >
> > fuse_main(argc, argv, &strFuseFuncSet, NULL);
> >
> > ...
> > }
> >
> > If I start my whole application 'fs_net' , where fuse is only a small part inside,
> > then my application has stopped after the call of fuse_main(). That
> > is the first think what is wrong.
>
> No, fuse_main just daemonizes your program. It is still running, and
> after you unmount the file system, fuse_main will return and program
> execution will continue normally.
>
>
> Best,
>
> -Nikolaus
>
> --
> »Time flies like an arrow, fruit flies like a Banana.«
>
> PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C
>
>
> ------------------------------------------------------------------------------
> Return on Information:
> Google Enterprise Search pays you back
> Get the facts.
> http://p.sf.net/sfu/google-dev2dev
> _______________________________________________
> fuse-devel mailing list
> fuse-devel@...
> https://lists.sourceforge.net/lists/listinfo/fuse-devel
>

Re: fuse example 'fusexmp.c' doesn't work as download path for a browser

2009-12-11T01:17:47Z

Hi Nikolaus,
that is right. With help of Sven I got the information that
the fusexmp.c mirrors the root directory of my linux and therefore
is filing the mountdir with the contents from my root-dir and
I got then also the root rights for the mountpoint.

Best Regards,
Maik

> -----Ursprüngliche Nachricht-----
> Von: "Nikolaus Rath" <Nikolaus@...>
> Gesendet: 11.12.09 00:13:23
> An: fuse-devel@...
> Betreff: Re: [fuse-devel] fuse example 'fusexmp.c' doesn't work as download path for a browser

> Sven Utcke <utcke@...> writes:
> > Hello Nikolaus,
> >
> >> >> > user1@UbuntuVM:~/workspace/fuse/fusetest$ ./fusexmp Fusexmp/
> >> >> > user1@UbuntuVM:~/workspace/fuse/fusetest$ ls -l
> >> >> > insgesamt 1752
> >> >> > -rwxr-xr-x 1 user1 user1 15565 2009-12-04 10:19 fusexmp
> >> >> > drwxr-xr-x 21 root root 4096 2009-10-20 14:22 Fusexmp
> >> >> > -rw-r--r-- 1 user1 user1 9646 2009-12-07 11:03 fusexmp.c
> >> >>
> >> >> Which uid and gid does the getattr() function of your file system return
> >> >> for '/'?
> >> >
> >> > The results are:
> >> >
> >> > call: xmp_getattr
> >> > call: ------> path: '/'
> >> > call: uid: '0'
> >> > call: gid: '0'
> >> >
> >> > But why I get the root rights for the mountdir
> >> > after starting of fuse?
> >>
> >> Because you told FUSE to do so. If you don't want the directory to be
> >> owned by root, don't return uid=gid=0 for it but the UID of the user
> >> that should be owner of the directory.
> >>
> >> Wild guess: You are aware that '/' refers to the root of *your*
> >> filesystem, i.e. the mountpoint, and not the real root directory of your
> >> system, aren't you?
> >
> > Well, no, actually it isn't, nor is he. Maik is using an essentially
> > unmodified fusexmp.c, which simply mirrors the root, so that '/' does
> > indeed refer to the systems root filesystem.
>
> He is? I thought he had modified it by now, because I remember him
> complaining that he had files in his mountpoint with a freshly mounted
> fs...
>
>
>
> Best,
>
>
> -Nikolaus
>
> --
> »Time flies like an arrow, fruit flies like a Banana.«
>
> PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C
>
>
> ------------------------------------------------------------------------------
> Return on Information:
> Google Enterprise Search pays you back
> Get the facts.
> http://p.sf.net/sfu/google-dev2dev
> _______________________________________________
> fuse-devel mailing list
> fuse-devel@...
> https://lists.sourceforge.net/lists/listinfo/fuse-devel
>

___________________________________________________________
Preisknaller: WEB.DE DSL Flatrate für nur 16,99 Euro/mtl.!
http://produkte.web.de/go/02/

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: fuse example 'fusexmp.c' doesn't work as download path for a browser

2009-12-10T13:38:17Z

Sven Utcke <utcke@...> writes:

> Hello Nikolaus,
>
>> >> > user1@UbuntuVM:~/workspace/fuse/fusetest$ ./fusexmp Fusexmp/
>> >> > user1@UbuntuVM:~/workspace/fuse/fusetest$ ls -l
>> >> > insgesamt 1752
>> >> > -rwxr-xr-x 1 user1 user1 15565 2009-12-04 10:19 fusexmp
>> >> > drwxr-xr-x 21 root root 4096 2009-10-20 14:22 Fusexmp
>> >> > -rw-r--r-- 1 user1 user1 9646 2009-12-07 11:03 fusexmp.c
>> >>
>> >> Which uid and gid does the getattr() function of your file system return
>> >> for '/'?
>> >
>> > The results are:
>> >
>> > call: xmp_getattr
>> > call: ------> path: '/'
>> > call: uid: '0'
>> > call: gid: '0'
>> >
>> > But why I get the root rights for the mountdir
>> > after starting of fuse?
>>
>> Because you told FUSE to do so. If you don't want the directory to be
>> owned by root, don't return uid=gid=0 for it but the UID of the user
>> that should be owner of the directory.
>>
>> Wild guess: You are aware that '/' refers to the root of *your*
>> filesystem, i.e. the mountpoint, and not the real root directory of your
>> system, aren't you?
>
> Well, no, actually it isn't, nor is he. Maik is using an essentially
> unmodified fusexmp.c, which simply mirrors the root, so that '/' does
> indeed refer to the systems root filesystem.

He is? I thought he had modified it by now, because I remember him
complaining that he had files in his mountpoint with a freshly mounted
fs...

Best,

-Nikolaus

--
»Time flies like an arrow, fruit flies like a Banana.«

PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: fuse example 'fusexmp.c' doesn't work as download path for a browser

2009-12-10T13:02:15Z

Hello Nikolaus,

> >> > user1@UbuntuVM:~/workspace/fuse/fusetest$ ./fusexmp Fusexmp/
> >> > user1@UbuntuVM:~/workspace/fuse/fusetest$ ls -l
> >> > insgesamt 1752
> >> > -rwxr-xr-x 1 user1 user1 15565 2009-12-04 10:19 fusexmp
> >> > drwxr-xr-x 21 root root 4096 2009-10-20 14:22 Fusexmp
> >> > -rw-r--r-- 1 user1 user1 9646 2009-12-07 11:03 fusexmp.c
> >>
> >> Which uid and gid does the getattr() function of your file system return
> >> for '/'?
> >
> > The results are:
> >
> > call: xmp_getattr
> > call: ------> path: '/'
> > call: uid: '0'
> > call: gid: '0'
> >
> > But why I get the root rights for the mountdir
> > after starting of fuse?
>
> Because you told FUSE to do so. If you don't want the directory to be
> owned by root, don't return uid=gid=0 for it but the UID of the user
> that should be owner of the directory.
>
> Wild guess: You are aware that '/' refers to the root of *your*
> filesystem, i.e. the mountpoint, and not the real root directory of your
> system, aren't you?

Well, no, actually it isn't, nor is he. Maik is using an essentially
unmodified fusexmp.c, which simply mirrors the root, so that '/' does
indeed refer to the systems root filesystem.

Sven
--
_ ___ ___ ___
__| |/ __|| __|/ __| The dCache File System
/ _` | (__ | _| \__ \ An archive file-system for PB of data
\__,_|\___||_| |___/ http://www.desy.de/~utcke/Data/

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: Fuse eats to much memory

2009-12-10T11:28:45Z

Can anybody help me? Is it usual behavior of fuse or not?

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: fuse example 'fusexmp.c' doesn't work as download path for a browser

2009-12-10T05:35:26Z

Hi,

Please consider to write your reply *below* the quote, deleting the
parts that you are not answering to.

Mylius WTW <Mylius_M@...> writes:

>> > That should actually not the reason, because I start fuse
>> > as (e.g) user1.
>> >
>> > 'Fusexmp' is my fuse mountpoint. Before I start fuse
>> > is that the 'ls -l' print of my fuse folder.
>> >
>> > user1@UbuntuVM:~/workspace/fuse/fusetest$ ls -l
>> > insgesamt 1752
>> > -rwxr-xr-x 1 user1 user1 15565 2009-12-04 10:19 fusexmp
>> > drwxr-xr-x 2 user1user1 4096 2009-12-04 09:41 Fusexmp
>> > -rw-r--r-- 1 user1 user1 9646 2009-12-07 11:03 fusexmp.c
>> >
>> > Nothing is root.
>> > And after starting of fuse, the mountpoint 'Fusexmp' has root rights:
>> >
>> > user1@UbuntuVM:~/workspace/fuse/fusetest$ ./fusexmp Fusexmp/
>> > user1@UbuntuVM:~/workspace/fuse/fusetest$ ls -l
>> > insgesamt 1752
>> > -rwxr-xr-x 1 user1 user1 15565 2009-12-04 10:19 fusexmp
>> > drwxr-xr-x 21 root root 4096 2009-10-20 14:22 Fusexmp
>> > -rw-r--r-- 1 user1 user1 9646 2009-12-07 11:03 fusexmp.c
>> >
>> >
>> > Strange .... . ?
>>
>> Which uid and gid does the getattr() function of your file system return
>> for '/'?
>
> The results are:
>
> call: xmp_getattr
> call: ------> path: '/'
> call: uid: '0'
> call: gid: '0'
>
> But why I get the root rights for the mountdir
> after starting of fuse?

Because you told FUSE to do so. If you don't want the directory to be
owned by root, don't return uid=gid=0 for it but the UID of the user
that should be owner of the directory.

Wild guess: You are aware that '/' refers to the root of *your*
filesystem, i.e. the mountpoint, and not the real root directory of your
system, aren't you?

Best,

-Nikolaus

--
»Time flies like an arrow, fruit flies like a Banana.«

PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: fuse example 'fusexmp.c' doesn't work as download path for a browser

2009-12-10T03:55:20Z

Hello Maik,

> user1 user1@UbuntuVM:~/workspace/fuse/fusetest/test$ ls -l
> ...
> -rwxr-xr-x 1 user1 user1 15565 2009-12-10 11:19 fusexmp
> drwxrwxrwx 2 user1 user1 4096 2009-12-10 10:37 Fusexmp
> -rw-r--r-- 1 user1 user1 9878 2009-12-10 11:13 fusexmp.c
> ...
>
> user1@UbuntuVM:~/workspace/fuse/fusetest/test$ ./fusexmp Fusexmp/
> user1@UbuntuVM:~/workspace/fuse/fusetest/test$ ls -l
> ...
> -rwxr-xr-x 1 user1 user1 15565 2009-12-10 11:19 fusexmp
> drwxr-xr-x 21 root root 4096 2009-10-20 14:22 Fusexmp
> -rw-r--r-- 1 user1 user1 9878 2009-12-10 11:13 fusexmp.c

Well, what did you expect? Fusexmp/ is now a mirror of /, which is
customarily owned by root...

> The results are:
>
> call: xmp_getattr
> call: ------> path: '/'
> call: uid: '0'
> call: gid: '0'

See, you are stating '/' --- not Fusexmp/, but '/' which just happens
to get mirrored to Fusexmp/ --- including all the original stat
information like rights and date (you noticed that changed too, didn't
you?).

> WIth the try to townload a file in the fuse mountdir
> I got the error messages:
>
> ERROR: with lstat() in xmp_getattr
> ERROR: with ------> path: '/downloadfile.jpg'
> ERROR: with errno: 'No such file or directory'

Sure --- even Fuse can not change the fact that you are not allowed to
write into the root as non-root user...

Does this help? Shameless plug: would it be a help to rent me (or
some of the others) some more and have this discussion in private?

Sven
--
_ ___ ___ ___
__| |/ __|| __|/ __| The dCache File System
/ _` | (__ | _| \__ \ An archive file-system for PB of data
\__,_|\___||_| |___/ http://www.desy.de/~utcke/Data/

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: fuse example 'fusexmp.c' doesn't work as download path for a browser

2009-12-10T02:49:10Z

The problem will be that after start of fuse the mountdir 'Fusexmp'
is root. And if I try to download a file with a browser with user right
then it is not possible to save the file inside the mountpoint, because
only root has then the rights to write inside the file.

user1 user1@UbuntuVM:~/workspace/fuse/fusetest/test$ ls -l
...
-rwxr-xr-x 1 user1 user1 15565 2009-12-10 11:19 fusexmp
drwxrwxrwx 2 user1 user1 4096 2009-12-10 10:37 Fusexmp
-rw-r--r-- 1 user1 user1 9878 2009-12-10 11:13 fusexmp.c
...

user1@UbuntuVM:~/workspace/fuse/fusetest/test$ ./fusexmp Fusexmp/
user1@UbuntuVM:~/workspace/fuse/fusetest/test$ ls -l
...
-rwxr-xr-x 1 user1 user1 15565 2009-12-10 11:19 fusexmp
drwxr-xr-x 21 root root 4096 2009-10-20 14:22 Fusexmp
-rw-r--r-- 1 user1 user1 9878 2009-12-10 11:13 fusexmp.c
...

I habe print the stat uid and gid values at start up if fuse with:

static int _getattr(const char *path, struct stat *stbuf)
{
int res;
char sbBuff[4096];

strcpy(sbBuff, path);

_fs_FU_logInfo("xmp_getattr");

sprintf(sbBuff, "---XXXXXX---> path: '%s'", path);
_fs_FU_logInfo(sbBuff);

res = lstat(path, stbuf);

if (res == -1)
{
_fs_FU_errorInfo("lstat() in _getattr");
_fs_FU_errorInfo(sbBuff);
sprintf(sbBuff, "errno: '%s'", strerror(errno));
_fs_FU_errorInfo(sbBuff);
return -errno;
}

sprintf(sbBuff, "uid: '%d'", stbuf->st_uid);
_fs_FU_logInfo(sbBuff);

sprintf(sbBuff, "gid: '%d'", stbuf->st_gid);
_fs_FU_logInfo(sbBuff);

return 0;
}

The results are:

call: xmp_getattr
call: ------> path: '/'
call: uid: '0'
call: gid: '0'

WIth the try to townload a file in the fuse mountdir
I got the error messages:

ERROR: with lstat() in xmp_getattr
ERROR: with ------> path: '/downloadfile.jpg'
ERROR: with errno: 'No such file or directory'

But I guess the is because of the root rigths.
But why I get the root rights for the mountdir
after starting of fuse?

user1@UbuntuVM:~/workspace/fuse/fusetest/test$ fusermount -u Fusexmp
user1@UbuntuVM:~/workspace/fuse/fusetest/test$ ls -l
-rwxr-xr-x 1 user1 user1 15601 2009-12-10 11:43 fusexmp
drwxrwxrwx 2 user1 user1 4096 2009-12-10 10:37 Fusexmp
-rw-r--r-- 1 user1 user1 9961 2009-12-10 11:44 fusexmp.c

Maik

> -----Ursprüngliche Nachricht-----
> Von: "Nikolaus Rath" <Nikolaus@...>
> Gesendet: 09.12.09 17:40:13
> An: fuse-devel@...
> Betreff: Re: [fuse-devel] fuse example 'fusexmp.c' doesn't work as download path for a browser

> Mylius WTW <Mylius_M@...> writes:
> > That should actually not the reason, because I start fuse
> > as (e.g) user1.
> >
> > 'Fusexmp' is my fuse mountpoint. Before I start fuse
> > is that the 'ls -l' print of my fuse folder.
> >
> > user1@UbuntuVM:~/workspace/fuse/fusetest$ ls -l
> > insgesamt 1752
> > -rwxr-xr-x 1 user1 user1 15565 2009-12-04 10:19 fusexmp
> > drwxr-xr-x 2 user1user1 4096 2009-12-04 09:41 Fusexmp
> > -rw-r--r-- 1 user1 user1 9646 2009-12-07 11:03 fusexmp.c
> >
> > Nothing is root.
> > And after starting of fuse, the mountpoint 'Fusexmp' has root rights:
> >
> > user1@UbuntuVM:~/workspace/fuse/fusetest$ ./fusexmp Fusexmp/
> > user1@UbuntuVM:~/workspace/fuse/fusetest$ ls -l
> > insgesamt 1752
> > -rwxr-xr-x 1 user1 user1 15565 2009-12-04 10:19 fusexmp
> > drwxr-xr-x 21 root root 4096 2009-10-20 14:22 Fusexmp
> > -rw-r--r-- 1 user1 user1 9646 2009-12-07 11:03 fusexmp.c
> >
> >
> > Strange .... . ?
>
> Which uid and gid does the getattr() function of your file system return
> for '/'?
>
>
>
>
> -Nikolaus
>
> --
> »Time flies like an arrow, fruit flies like a Banana.«
>
> PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C
>
>
> ------------------------------------------------------------------------------
> Return on Information:
> Google Enterprise Search pays you back
> Get the facts.
> http://p.sf.net/sfu/google-dev2dev
> _______________________________________________
> fuse-devel mailing list
> fuse-devel@...
> https://lists.sourceforge.net/lists/listinfo/fuse-devel
>

Re: fuse path encodings

2009-12-09T16:03:31Z

Ok thanks guys, this last round has cleared things up I think. I guess the
requirement about no '\0's just limits the encodings you can use to 8-bit
safe ones. I have to say I've never seen a system in a UTF-16 locale. My
filesystem doesn't care what the byte sequences mean, and it sounds like I'm
safe to go on handling them opaquely.

Another thought that hit me is that filesystems like mine that are mainly
used interactively (e.g. browsed with a GUI, or with "ls; cd; ls; cd") will
likely only ever get requests for "/", and thereafter paths that they
themselves have returned from readdir() anyway.

Thanks,

matt
------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: call of fuse_main() in a application

2009-12-09T10:08:49Z

Mylius WTW <Mylius_M@...> writes:

> Hallo,
>
> I want to call the function in a application where
> a lot of other threads are also running. If I call the
> fuse_main function with static parameters like:
>
> func_xyz()
> {
> ...
> int argc = 2;
> char *argv[] =
> {
> "./fs_net",
> "mountdir"
> };
>
> fuse_main(argc, argv, &strFuseFuncSet, NULL);
>
> ...
> }
>
> If I start my whole application 'fs_net' , where fuse is only a small part inside,
> then my application has stopped after the call of fuse_main(). That
> is the first think what is wrong.

No, fuse_main just daemonizes your program. It is still running, and
after you unmount the file system, fuse_main will return and program
execution will continue normally.

Best,

-Nikolaus

--
»Time flies like an arrow, fruit flies like a Banana.«

PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: fuse example 'fusexmp.c' doesn't work as download path for a browser

2009-12-09T07:04:37Z

Mylius WTW <Mylius_M@...> writes:

> That should actually not the reason, because I start fuse
> as (e.g) user1.
>
> 'Fusexmp' is my fuse mountpoint. Before I start fuse
> is that the 'ls -l' print of my fuse folder.
>
> user1@UbuntuVM:~/workspace/fuse/fusetest$ ls -l
> insgesamt 1752
> -rwxr-xr-x 1 user1 user1 15565 2009-12-04 10:19 fusexmp
> drwxr-xr-x 2 user1user1 4096 2009-12-04 09:41 Fusexmp
> -rw-r--r-- 1 user1 user1 9646 2009-12-07 11:03 fusexmp.c
>
> Nothing is root.
> And after starting of fuse, the mountpoint 'Fusexmp' has root rights:
>
> user1@UbuntuVM:~/workspace/fuse/fusetest$ ./fusexmp Fusexmp/
> user1@UbuntuVM:~/workspace/fuse/fusetest$ ls -l
> insgesamt 1752
> -rwxr-xr-x 1 user1 user1 15565 2009-12-04 10:19 fusexmp
> drwxr-xr-x 21 root root 4096 2009-10-20 14:22 Fusexmp
> -rw-r--r-- 1 user1 user1 9646 2009-12-07 11:03 fusexmp.c
>
>
> Strange .... . ?

Which uid and gid does the getattr() function of your file system return
for '/'?

-Nikolaus

--
»Time flies like an arrow, fruit flies like a Banana.«

PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

call of fuse_main() in a application

2009-12-09T05:25:26Z

Hallo,

I want to call the function in a application where
a lot of other threads are also running. If I call the
fuse_main function with static parameters like:

func_xyz()
{
...
int argc = 2;
char *argv[] =
{
"./fs_net",
"mountdir"
};

fuse_main(argc, argv, &strFuseFuncSet, NULL);

...
}

If I start my whole application 'fs_net' , where fuse is only a small part inside,
then my application has stopped after the call of fuse_main(). That
is the first think what is wrong.

If I start the application afterwards again, then I got the information:
'mount: from mtab, mountdir is already mounted ...'
(german: mount: Laut mtab ist xxx schon auf /root/workspace/fs_net/mountdir eingehängt)
and the application has not stopped, is running continuing.

The mountpoint 'mountdir' is not to see on the desktop, where he otherwise appears.
But the folder is mounted:
user1@UbuntuVM:~$ cat /proc/mounts | grep fs_net
fs_net /root/workspace/fs_net/mountdir fuse.fs_net rw,nosuid,nodev,user_id=0,group_id=0 0 0

How can I correct call the fuse_main() function inside a other application?
And is the first parameter in argv (argv[0]) important, or it is only a string where
the mountpoint in mtab is dedicted?

Maik

______________________________________________________
GRATIS für alle WEB.DE-Nutzer: Die maxdome Movie-FLAT!
Jetzt freischalten unter http://movieflat.web.de

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: fuse example 'fusexmp.c' doesn't work as download path for a browser

2009-12-08T23:23:00Z

That should actually not the reason, because I start fuse
as (e.g) user1.

'Fusexmp' is my fuse mountpoint. Before I start fuse
is that the 'ls -l' print of my fuse folder.

user1@UbuntuVM:~/workspace/fuse/fusetest$ ls -l
insgesamt 1752
-rwxr-xr-x 1 user1 user1 15565 2009-12-04 10:19 fusexmp
drwxr-xr-x 2 user1user1 4096 2009-12-04 09:41 Fusexmp
-rw-r--r-- 1 user1 user1 9646 2009-12-07 11:03 fusexmp.c

Nothing is root.
And after starting of fuse, the mountpoint 'Fusexmp' has root rights:

user1@UbuntuVM:~/workspace/fuse/fusetest$ ./fusexmp Fusexmp/
user1@UbuntuVM:~/workspace/fuse/fusetest$ ls -l
insgesamt 1752
-rwxr-xr-x 1 user1 user1 15565 2009-12-04 10:19 fusexmp
drwxr-xr-x 21 root root 4096 2009-10-20 14:22 Fusexmp
-rw-r--r-- 1 user1 user1 9646 2009-12-07 11:03 fusexmp.c

And after the umount: ' fusermount -u Fusexmp/' has the mountdir
'Fusexmp' user rigths again:

user1@UbuntuVM:~/workspace/fuse/fusetest$ fusermount -u Fusexmp/
user1@UbuntuVM:~/workspace/fuse/fusetest$ ls -l
insgesamt 1752
-rwxr-xr-x 1 user1 user1 15565 2009-12-04 10:19 fusexmp
drwxr-xr-x 2 user1 user1 4096 2009-12-04 09:41 Fusexmp
-rw-r--r-- 1 user1 user1 9646 2009-12-07 11:03 fusexmp.c

Strange .... . ?

Maik

> -----Ursprüngliche Nachricht-----
> Von: "Nikolaus Rath" <Nikolaus@...>
> Gesendet: 08.12.09 18:38:53
> An: fuse-devel@...
> Betreff: Re: [fuse-devel] fuse example 'fusexmp.c' doesn't work as download path for a browser

> My guess is that you are starting fuse as root without --allow_other and
> then try to access the mountpoint as a normal user.
>
> Mylius WTW <Mylius_M@...> writes:
> > Hi Nikolaus,
> >
> > that what you have written could be the reason.
> > I have tried to copy a file on the command line from outside
> > inside the mountpoint and it wasn't possible. Before I have
> > started the fuse program belongs the dir (mountpoint afterwards)
> > the userX. But if I launch the fuse program with the mountpoint
> > belongs the mountdir 'root' and not more userX. Therfore I have as
> > userX no permission more to write the file inside.
> >
> > I must now found out what is the reason .... .
> >
> > Maik
> >
> >
> >
> >
> >> -----Ursprüngliche Nachricht-----
> >> Von: "Nikolaus Rath" <Nikolaus@...>
> >> Gesendet: 07.12.09 16:43:08
> >> An: fuse-devel@...
> >> Betreff: Re: [fuse-devel] fuse example 'fusexmp.c' doesn't work as download path for a browser
> >
> >
> >> Mylius WTW <Mylius_M@...> writes:
> >> > Hi,
> >> >
> >> > I have taken follow fuse example: 'fusexmp.c'
> >> >
> >> > http://fuse.cvs.sourceforge.net/viewvc/*checkout*/fuse/fuse/example/fusexmp.c?revision=1.40
> >> >
> >> > Then I have taken the fuse mountpoint als path for the browser to download a file.
> >> > Then if I download a file, the download windows starts but no download is happend.
> >> >
> >> > Follow functions are called if I try a download:
> >> >
> >> > call: xmp_getattr
> >> > call: xmp_mknod
> >> > call: xmp_access
> >> > call: xmp_getattr
> >> > call: xmp_mknod
> >> >
> >> >
> >> > And I get e.g. in function 'xmp_getattr()'
> >> > with call of 'lstat()' the error return value of -1 back:
> >> > the 'path' containts: '/download_file.mp3'
> >> >
> >> > Should the example 'fusexmp.c' work correct?
> >>
> >> Can you copy the problematic file inside the mount point with cp on the
> >> command line? Can you copy the original file outside the mountpoint
> >> (fusexmp.c just mirrors an existing directory)?
> >>
> >>
> >> Best,
> >>
> >> -Nikolaus
> >>
> >> --
> >> »Time flies like an arrow, fruit flies like a Banana.«
> >>
> >> PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C
> >>
> >>
> >> ------------------------------------------------------------------------------
> >> Join us December 9, 2009 for the Red Hat Virtual Experience,
> >> a free event focused on virtualization and cloud computing.
> >> Attend in-depth sessions from your desk. Your couch. Anywhere.
> >> http://p.sf.net/sfu/redhat-sfdev2dev
> >> _______________________________________________
> >> fuse-devel mailing list
> >> fuse-devel@...
> >> https://lists.sourceforge.net/lists/listinfo/fuse-devel
> >>
> >
> >
> > ______________________________________________________
> > GRATIS für alle WEB.DE-Nutzer: Die maxdome Movie-FLAT!
> > Jetzt freischalten unter http://movieflat.web.de
> >
> >
> > ------------------------------------------------------------------------------
> > Return on Information:
> > Google Enterprise Search pays you back
> > Get the facts.
> > http://p.sf.net/sfu/google-dev2dev
>
>
> -Nikolaus
>
> --
> »Time flies like an arrow, fruit flies like a Banana.«
>
> PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C
>
>
> ------------------------------------------------------------------------------
> Return on Information:
> Google Enterprise Search pays you back
> Get the facts.
> http://p.sf.net/sfu/google-dev2dev
> _______________________________________________
> fuse-devel mailing list
> fuse-devel@...
> https://lists.sourceforge.net/lists/listinfo/fuse-devel
>

Re: fuse example 'fusexmp.c' doesn't work as download path for a browser

2009-12-08T07:24:39Z

My guess is that you are starting fuse as root without --allow_other and
then try to access the mountpoint as a normal user.

Mylius WTW <Mylius_M@...> writes:

> Hi Nikolaus,
>
> that what you have written could be the reason.
> I have tried to copy a file on the command line from outside
> inside the mountpoint and it wasn't possible. Before I have
> started the fuse program belongs the dir (mountpoint afterwards)
> the userX. But if I launch the fuse program with the mountpoint
> belongs the mountdir 'root' and not more userX. Therfore I have as
> userX no permission more to write the file inside.
>
> I must now found out what is the reason .... .
>
> Maik
>
>
>
>
>> -----Ursprüngliche Nachricht-----
>> Von: "Nikolaus Rath" <Nikolaus@...>
>> Gesendet: 07.12.09 16:43:08
>> An: fuse-devel@...
>> Betreff: Re: [fuse-devel] fuse example 'fusexmp.c' doesn't work as download path for a browser
>
>
>> Mylius WTW <Mylius_M@...> writes:
>> > Hi,
>> >
>> > I have taken follow fuse example: 'fusexmp.c'
>> >
>> > http://fuse.cvs.sourceforge.net/viewvc/*checkout*/fuse/fuse/example/fusexmp.c?revision=1.40
>> >
>> > Then I have taken the fuse mountpoint als path for the browser to download a file.
>> > Then if I download a file, the download windows starts but no download is happend.
>> >
>> > Follow functions are called if I try a download:
>> >
>> > call: xmp_getattr
>> > call: xmp_mknod
>> > call: xmp_access
>> > call: xmp_getattr
>> > call: xmp_mknod
>> >
>> >
>> > And I get e.g. in function 'xmp_getattr()'
>> > with call of 'lstat()' the error return value of -1 back:
>> > the 'path' containts: '/download_file.mp3'
>> >
>> > Should the example 'fusexmp.c' work correct?
>>
>> Can you copy the problematic file inside the mount point with cp on the
>> command line? Can you copy the original file outside the mountpoint
>> (fusexmp.c just mirrors an existing directory)?
>>
>>
>> Best,
>>
>> -Nikolaus
>>
>> --
>> »Time flies like an arrow, fruit flies like a Banana.«
>>
>> PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C
>>
>>
>> ------------------------------------------------------------------------------
>> Join us December 9, 2009 for the Red Hat Virtual Experience,
>> a free event focused on virtualization and cloud computing.
>> Attend in-depth sessions from your desk. Your couch. Anywhere.
>> http://p.sf.net/sfu/redhat-sfdev2dev
>> _______________________________________________
>> fuse-devel mailing list
>> fuse-devel@...
>> https://lists.sourceforge.net/lists/listinfo/fuse-devel
>>
>
>
> ______________________________________________________
> GRATIS für alle WEB.DE-Nutzer: Die maxdome Movie-FLAT!
> Jetzt freischalten unter http://movieflat.web.de
>
>
> ------------------------------------------------------------------------------
> Return on Information:
> Google Enterprise Search pays you back
> Get the facts.
> http://p.sf.net/sfu/google-dev2dev

Re: fuse path encodings

2009-12-08T01:57:46Z

> A file name for the kernel and FUSE is a sequence of bytes, not a
> sequence of characters. There are some limitations on the sequence of
> bytes: it has a maximum length (255),

Is this a kernel limitation? I don't believe it is a limitation in
FUSE --- does this mean I ought to check for names longer than 255
Characters in my implementation, or else something might break in the
kernel?

Sven
--
_ ___ ___ ___
__| |/ __|| __|/ __| The dCache File System
/ _` | (__ | _| \__ \ An archive file-system for PB of data
\__,_|\___||_| |___/ http://www.desy.de/~utcke/Data/

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: after launch of fuse are a lot of files and dirs inside the mountpoint

2009-12-08T01:48:58Z

Hi Sven,

that is rigth. That should be the reason. I try now
to solve this problem.

Must I hand over a second mountdir which is empty
the function 'fuse_main'

like

...
fuse_main(argc, argv, strFuncSet, XX);
...

and then launch the program with e.g.:

'./fuse emtydir mountdir'

?

Is such a way possible ?

Maik

> -----Ursprüngliche Nachricht-----
> Von: "Sven Utcke" <utcke@...>
> Gesendet: 04.12.09 16:58:38
> An: fuse-devel@...
> Betreff: Re: [fuse-devel] after launch of fuse are a lot of files and dirs inside the mountpoint

> > That is right,
> >
> > in function 'xmp_readdir ' comes mirrored the root dir.
> >
> > The handed over 'path' string contains the root-dir '/'.
> >
> > But what is the reason that the root-path '/' is handed over
> > to fill afterwards with the filler method the mountpoint ?
>
> You get ""+path, so if your input is "/" that's what you get...
>
> > How can I get a empty mountpoint after launch of the fuse
> > filesystem?
>
> Rewrite all functions to use a prefix of your liking.
>
> Sven
> --
> _ ___ ___ ___
> __| |/ __|| __|/ __| The dCache File System
> / _` | (__ | _| \__ \ An archive file-system for PB of data
> \__,_|\___||_| |___/ http://www.desy.de/~utcke/Data/
>
> ------------------------------------------------------------------------------
> Join us December 9, 2009 for the Red Hat Virtual Experience,
> a free event focused on virtualization and cloud computing.
> Attend in-depth sessions from your desk. Your couch. Anywhere.
> http://p.sf.net/sfu/redhat-sfdev2dev
> _______________________________________________
> fuse-devel mailing list
> fuse-devel@...
> https://lists.sourceforge.net/lists/listinfo/fuse-devel
>

Re: fuse example 'fusexmp.c' doesn't work as download path for a browser

2009-12-08T01:21:28Z

Hi Nikolaus,

that what you have written could be the reason.
I have tried to copy a file on the command line from outside
inside the mountpoint and it wasn't possible. Before I have
started the fuse program belongs the dir (mountpoint afterwards)
the userX. But if I launch the fuse program with the mountpoint
belongs the mountdir 'root' and not more userX. Therfore I have as
userX no permission more to write the file inside.

I must now found out what is the reason .... .

Maik

> -----Ursprüngliche Nachricht-----
> Von: "Nikolaus Rath" <Nikolaus@...>
> Gesendet: 07.12.09 16:43:08
> An: fuse-devel@...
> Betreff: Re: [fuse-devel] fuse example 'fusexmp.c' doesn't work as download path for a browser

> Mylius WTW <Mylius_M@...> writes:
> > Hi,
> >
> > I have taken follow fuse example: 'fusexmp.c'
> >
> > http://fuse.cvs.sourceforge.net/viewvc/*checkout*/fuse/fuse/example/fusexmp.c?revision=1.40
> >
> > Then I have taken the fuse mountpoint als path for the browser to download a file.
> > Then if I download a file, the download windows starts but no download is happend.
> >
> > Follow functions are called if I try a download:
> >
> > call: xmp_getattr
> > call: xmp_mknod
> > call: xmp_access
> > call: xmp_getattr
> > call: xmp_mknod
> >
> >
> > And I get e.g. in function 'xmp_getattr()'
> > with call of 'lstat()' the error return value of -1 back:
> > the 'path' containts: '/download_file.mp3'
> >
> > Should the example 'fusexmp.c' work correct?
>
> Can you copy the problematic file inside the mount point with cp on the
> command line? Can you copy the original file outside the mountpoint
> (fusexmp.c just mirrors an existing directory)?
>
>
> Best,
>
> -Nikolaus
>
> --
> »Time flies like an arrow, fruit flies like a Banana.«
>
> PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C
>
>
> ------------------------------------------------------------------------------
> Join us December 9, 2009 for the Red Hat Virtual Experience,
> a free event focused on virtualization and cloud computing.
> Attend in-depth sessions from your desk. Your couch. Anywhere.
> http://p.sf.net/sfu/redhat-sfdev2dev
> _______________________________________________
> fuse-devel mailing list
> fuse-devel@...
> https://lists.sourceforge.net/lists/listinfo/fuse-devel
>

Re: fuse path encodings

2009-12-08T01:09:24Z

On 08/12/09 02:37, Nikolaus Rath wrote:
> No, you get all paths as byte sequences. It is up to you if you want to
> interpret them as encoded strings. For FUSE and the kernel they are just
> binary data with the added requirement that they must not contain \0 or
> /.
>

Absolutely. This is a point worth repeating.

A file name for the kernel and FUSE is a sequence of bytes, not a
sequence of characters. There are some limitations on the sequence of
bytes: it has a maximum length (255), it must not contain \0 and it must
not contain /. What anyone does with that sequence of bytes is entirely
up to them: the kernel neither knows nor cares -- you can choose to
interpret it as UTF-8, SJIS, ISO-8859-15, base64 UTF-16, whatever.

[Many years ago I implemented a help function for an editor by storing
the help for individual key sequences in files whose name were the key
sequences: for example, the help for ctrl-C was stored in a file whose
name was the single character '\003'. I think I had a separate program
to rename the files into something I could type; it saved a precious few
bytes on a PDP-11 doing it that way.]

jch

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: fuse path encodings

2009-12-07T18:37:13Z

Matt Turner <matturner@...> writes:

> 2009/12/4 Goswin von Brederlow <goswin-v-b@...>
>
>> John Haxby <john.haxby@...> writes:
>>
>> > On 03/12/09 22:58, Matt Turner wrote:
>> >> I was wondering about dealing with non-ascii characters in paths. Simply
>> >> put, in what encoding are paths passed to a libfuse client?
>> >>
>> >>
>> >
>> >
>> > Unless something has happened since I last looked (and I don't think it
>> > has) the kernel and fuse neither know nor care what the encoding used
>> > for a file name is. With two exceptions. The octets 0x2f (/) and 0x00
>> > are used to separate pathname components and terminate pathnames
>> > respectively. Of course, you would have trouble creating a file called
>> > "." or ".." (because they already exist) and if you had a file system
>> > that didn't have those all kinds of things would break in interesting
>> ways.
>>
> What does this mean for strings with 0x2f or 0x00 in them? For example,
> non-accented latin characters ("ascii"), like all the names in a unix
> installation, encoded in utf-16 is half 0x00 bytes. It it safe to pass these
> to the kernel?

No, you can't pass \0 to the kernel. That's why you usually use UTF-8,
which does not generate any problematic bytes.

> It obviously can't just use strlen() on them. Maybe it checks
> $LC_LANG etc and deals with different encodings?

No, the kernel does not care about the locale at all and does not know
anything about encodings.

> You say fuse neither knows nor cares about the encoding (although surely it
> must if it so much as strdup()s any strings).

This doesn't really have anything to do with encoding. You cannot have a
\0 in unix filenames, no matter how you would interpret it.

> Does this just mean I get all
> paths in the system's encoding (that returned by nl_langinfo(CODESET))? If
> this is guaranteed then coping with it is easy.

No, you get all paths as byte sequences. It is up to you if you want to
interpret them as encoded strings. For FUSE and the kernel they are just
binary data with the added requirement that they must not contain \0 or
/.

Best,

-Nikolaus

--
»Time flies like an arrow, fruit flies like a Banana.«

PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: Using my own inodes with FUSE

2009-12-07T18:31:27Z

Amruta Gokhale <amruta.gokhale@...> writes:
> As I understand it, FUSE uses and assigns its own inode numbers to the files
> and directories.
> And to override this and make FUSE use my inode implementation, I need to
> give "-o use_ino" option.

That is correct.

> Is there something special that I should do for using my own inodes
> implementation?
> Something like, do I need to implement functions in fuse__lowlevel__ops
> structure?

No, nothing like that is necessary.

> What is not working is my inode implementation.
> bash$mkdir tmp
> bash$./dfs -o use_ino tmp
> bash$ls -li
> ls: cannot access tmp: No such file or directory
> 2613945 d????????? ? ? ? ? ? tmp
> ^^^^^ This was the inode number assigned by linux FS to the "tmp" directory.
> I had assigned inode number "1" to the root directory. And I had other
> information
> stored in my inode. But none of this information is visible here.
>
> Whey I tried debugging, the getattr() function prints the correct
> inode information as I want, if I insert some print statements. I am
> unable to understand why this doesn't work then.

Since you are not able to understand why it does not work, I am
surprised that you still feel confident to judge if getattr() prints the
correct information.

If you want other people to help you, you really have to give them the
chance to look at the actual code and actual data rather than your
interpretations of them.

So I recommend that you either reduce your problem to a *short* example
file system that we can run and reproduce the problem ourselves (this is
always the best way to get help) or at least tell us *how* you debugged
the problem, *what* print statements you inserted *where*, and *what
exactly* they print when the fs is running.

> Is there an example filesystem source code available, which implements its
> own inodes and uses FUSE?

Not that I know of, at least not a simple example.

Best,

-Nikolaus

--
»Time flies like an arrow, fruit flies like a Banana.«

PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6 02CF A9AD B7F8 AE4E 425C

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: fuse path encodings

2009-12-07T16:39:59Z

2009/12/4 Goswin von Brederlow <goswin-v-b@...>

> John Haxby <john.haxby@...> writes:
>
> > On 03/12/09 22:58, Matt Turner wrote:
> >> I was wondering about dealing with non-ascii characters in paths. Simply
> >> put, in what encoding are paths passed to a libfuse client?
> >>
> >>
> >
> >
> > Unless something has happened since I last looked (and I don't think it
> > has) the kernel and fuse neither know nor care what the encoding used
> > for a file name is. With two exceptions. The octets 0x2f (/) and 0x00
> > are used to separate pathname components and terminate pathnames
> > respectively. Of course, you would have trouble creating a file called
> > "." or ".." (because they already exist) and if you had a file system
> > that didn't have those all kinds of things would break in interesting
> ways.
>

What does this mean for strings with 0x2f or 0x00 in them? For example,
non-accented latin characters ("ascii"), like all the names in a unix
installation, encoded in utf-16 is half 0x00 bytes. It it safe to pass these
to the kernel? It obviously can't just use strlen() on them. Maybe it checks
$LC_LANG etc and deals with different encodings?

You say fuse neither knows nor cares about the encoding (although surely it
must if it so much as strdup()s any strings). Does this just mean I get all
paths in the system's encoding (that returned by nl_langinfo(CODESET))? If
this is guaranteed then coping with it is easy.

> >
> > So, I can create a file called "\x88\x68", "\x68\x88", "\xe8\xa1\xa8" or
> > "\x95\x5c" (representations of U+8868 in UCS-2LE, UTF-16, UTF-8 and
> > SJIS) and how they look depends on the locale that the terminal is using
> > to interpret character representations. The last one is interesting
> > because it ends with 0x5c (\) but the file system doesn't much care
> > about that.
>
Obviously the normal rules apply about not using external strings as
printf() format arguments :) You're quite right that representation is
different from the byte stream. The symbol you get depends on the code point
(again, separate from the encoding) and the type face (most of which don't
cover all code points anyway). My filesystem doesn't care what the names
look like, or what they mean, so has no reason to decode a string to find
the codepoints it represents. I do need strchr() to work though.

> >
> > It is possible to create a file system that, for example, stores its
> > pathnames as vectors of utf16 strings but you would have to make a
> > decision about how to convert what the is passed into the kernel into
> > UTF-16 -- that would have to be a mount option or something that you
> > derive from somewhere (eg nl_langinfo(CODESET)) but it wouldn't stop
> > someone creating a file called "\x95\x5c" because they happen to want to
> > use SJIS but you were expecting UTF-8.
>
If I /know/ I'm getting utf-16 then I can deal with it. Either handle utf-16
natively, or convert to wchar_t on input with mbsrtowcs(), and then back to
utf-16 on output. I'm not sure how a mount option would help here - strings
will come in in whatever the system's encoding is, making the user specify
that seems redundant. Some filesystems do have an option which tells them
what to use _on disk_, and more importantly what's currently on the disk.
This could be different to both what comes in, and the program's internal
encoding.

As you say, if I find something in an unexpected encoding on disk there's
not much I can do. I'm not sure what would happen if another program ignored
LC_LANG and passed the wrong encoding to the kernel. I guess if it survived
the trip thorough it'd just get delivered to me.

> >
>
> I believe FAT has just such a mount option and will translate from the
> actual locale used to the one given as option.
>

The only option I can find for FAT is iocharset. A few FSes have this, it
seems to be which 8-bit *character set* to use when converting from unicode
on-disk to 8bit. I guess this only applies if your system is in an 8bit
locale.
------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel

Re: fuse: device not found, try 'modprobe fuse' first

2009-12-07T14:06:58Z

Michael Muratet wrote:

>
> On Dec 7, 2009, at 3:30 PM, Stephen Munnings wrote:
>> Michael Muratet wrote:
>>> I am trying to create an sshd file system, and I get the error
>>> "device not found try modprobe fuse first. I did and it does not
>>> find fuse. I tried the example in the example directory and it
>>> doesn't work either, same error. The code compiled and installed
>>> without so much as a warning.
>>>
>>> The kernel appears to be the correct vintage:
>>>
>>> scyld:example root$ uname -a
>>> Linux scyld.localdomain 2.6.18-128.1.1.el5.530g0000 #1 SMP Tue Mar
>>> 17 21:49:24 EDT 2009 x86_64 x86_64 x86_64 GNU/Linux
>> There needs to be a /dev/fuse on your system.
> There is. It's dated today and I believe it got there when I built fuse.

How was it built? What do you see when you do an "ls -l /dev/fuse"?
>> It is possible that the kernel was not compiled with FUSE support -
>> either as a built-in or as a loadable module.
>>
>> If the kernel was compiled with this as a built-in - /dev/fuse should
>> be there when you boot.
> Do I need to reboot the system to get the installation to be recognized?
Perhaps, I am not sure - especially with your setup. My experience with
Linux (little as it is) is more with embedded systems.
>> If the kernel was compiled with FUSE support as a loadable module,
>> then the "modprobe fuse" should install it.
> It's a beowulf cluster and the odds are good the kernel was not
> compiled with fuse support, but, I thought fuse's claim to fame was
> that you didn't have to mess with the kernel.
Well, you do need the basic support in the kernel, but once that is in,
all further FUSE development (in theory) need not touch kernel code at all.
It is made up of 3 parts (as far as I can tell)
1. The kernel support (this needs to be there when you start the
"application")
2. The libraries (to handle large amounts of common code and interfacing)
3. The actual application - that is the part you and I normally mess with.
> Thanks
>
> Mike
I am a newcomer to FUSE myself, so don't hang your hat too heavily on my
advice.

Regards,

Stephen Munnings

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
fuse-devel mailing list
fuse-devel@...
https://lists.sourceforge.net/lists/listinfo/fuse-devel