[PATCH vfat] allow retrieving entries with trailing dots

[PATCH vfat] allow retrieving entries with trailing dots

[Philippe De Muyter]

Hello Ogawa,

This fixes accessing vfat entries with trailing dots created by an external
vfat driver (like the one in IOMEGA home network hard drives)

Philippe

--

Some vfat-formatted network disks that are also usb disk do not discard
trailing dots when creating files or directories via ftp.
Connecting afterwards this drive via usb to a linux machine leads to the
following problem :
if one issues the `ls' or `find' command, one gets this message :

find: ./Simon_&_Garfunkel-Wednesday_Morning,_3_a.m.: No such file or directory

Fix that by first trying to retrieve the entry with the full name, and only if
that fails and there are trailing dots in the searched name, try then to find
the truncated name.

Signed-off-by: Philippe De Muyter <phdm@macqel.be>

--- a/fs/fat/namei_vfat.c	2009-09-10 00:13:59.000000000 +0200
+++ b/fs/fat/namei_vfat.c	2010-02-08 02:28:37.010096903 +0100
@@ -702,9 +702,22 @@
 static int vfat_find(struct inode *dir, struct qstr *qname,
 		     struct fat_slot_info *sinfo)
 {
-	unsigned int len = vfat_striptail_len(qname);
+	int err;
+	unsigned int len;
+
+	/* Some combined ethernet + usb external hard drive do not
+	 * remove the trailing dots when creating entries in ethernet mode.
+	 * (e.g. Iomega Home Network Hard Drive)
+	 * Make accessing those entries possible.
+	 */
+	err = fat_search_long(dir, qname->name, qname->len, sinfo);
+	if (!err)
+		return err;
+	len = vfat_striptail_len(qname);
 	if (len == 0)
 		return -ENOENT;
+	if (len == qname->len)
+		return err;
 	return fat_search_long(dir, qname->name, len, sinfo);
 }
 

Re: [PATCH vfat] allow retrieving entries with trailing dots

[OGAWA Hirofumi]

Philippe De Muyter <phdm@macqel.be> writes:

> --- a/fs/fat/namei_vfat.c	2009-09-10 00:13:59.000000000 +0200
> +++ b/fs/fat/namei_vfat.c	2010-02-08 02:28:37.010096903 +0100
> @@ -702,9 +702,22 @@
>  static int vfat_find(struct inode *dir, struct qstr *qname,
>  		     struct fat_slot_info *sinfo)
>  {
> -	unsigned int len = vfat_striptail_len(qname);
> +	int err;
> +	unsigned int len;
> +
> +	/* Some combined ethernet + usb external hard drive do not
> +	 * remove the trailing dots when creating entries in ethernet mode.
> +	 * (e.g. Iomega Home Network Hard Drive)
> +	 * Make accessing those entries possible.
> +	 */
> +	err = fat_search_long(dir, qname->name, qname->len, sinfo);
> +	if (!err)
> +		return err;
> +	len = vfat_striptail_len(qname);
>  	if (len == 0)
>  		return -ENOENT;
> +	if (len == qname->len)
> +		return err;
>  	return fat_search_long(dir, qname->name, len, sinfo);
>  }

This would be bad for both (standard and IO-MEGA hack).

This introduces unneeded directory-parse to standard one.  And for
IO-MEGA, this wouldn't provide proper filename handling.

If it wants to handle the tailing-dot as a part of filename, it
shouldn't be able to access to the stripped-dots filename. (For simple
example, I guess you can't do "mv a a." with this patch.)

I'm not still sure if we should include this or not, however, I think
improper/imperfect quick hack wouldn't be acceptable at least.

Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: [PATCH vfat] allow retrieving entries with trailing dots

[Philippe De Muyter]

Hello Ogawa

Thanks for the quick answer.

On Wed, Mar 10, 2010 at 11:44:27PM +0900, OGAWA Hirofumi wrote:
> Philippe De Muyter <phdm@macqel.be> writes:
> 
> > --- a/fs/fat/namei_vfat.c	2009-09-10 00:13:59.000000000 +0200
> > +++ b/fs/fat/namei_vfat.c	2010-02-08 02:28:37.010096903 +0100
> > @@ -702,9 +702,22 @@
> >  static int vfat_find(struct inode *dir, struct qstr *qname,
> >  		     struct fat_slot_info *sinfo)
> >  {
> > -	unsigned int len = vfat_striptail_len(qname);
> > +	int err;
> > +	unsigned int len;
> > +
> > +	/* Some combined ethernet + usb external hard drive do not
> > +	 * remove the trailing dots when creating entries in ethernet mode.
> > +	 * (e.g. Iomega Home Network Hard Drive)
> > +	 * Make accessing those entries possible.
> > +	 */
> > +	err = fat_search_long(dir, qname->name, qname->len, sinfo);
> > +	if (!err)
> > +		return err;
> > +	len = vfat_striptail_len(qname);
> >  	if (len == 0)
> >  		return -ENOENT;
> > +	if (len == qname->len)
> > +		return err;
> >  	return fat_search_long(dir, qname->name, len, sinfo);
> >  }
> 
> This would be bad for both (standard and IO-MEGA hack).
> 
> This introduces unneeded directory-parse to standard one.  And for

No.  There will only be a second parse if someone is looking for a filename
with a trailing dot, and it is not found (*).  I there is no trailing dot in qname, only the first fat_search_long will be called, because len after vfat_striptail_len would be equal to qname->len.

> IO-MEGA, this wouldn't provide proper filename handling.

For accessing IO-MEGA disk in read mode, this is perfect.  I didn't want
to replicate IO-MEGA write behaviour here, only fix the read behaviour for
such simple commands as 'ls', 'find' and all the directory browsers.

> 
> If it wants to handle the tailing-dot as a part of filename, it
> shouldn't be able to access to the stripped-dots filename. (For simple
> example, I guess you can't do "mv a a." with this patch.)

As I explained above, I only fix read-access on IO-MEGA drives, while
preserving standard behaviour for write mode.

But I'll try your testcase asap.  Which behaviour do you expect ?
I would expect a no-op, because I did not change the write-behaviour.

Best regards

Philippe

(*) This will never happen with 'ls', 'find' and friends who get their name
list from getdents.  This could only happen if someone tries to open
a file called 'a.' on his vfat file-system, which of course does not exist
in his normal vfat file-system.

Re: [PATCH vfat] allow retrieving entries with trailing dots

[OGAWA Hirofumi]

Philippe De Muyter <phdm@macqel.be> writes:

>> This introduces unneeded directory-parse to standard one.  And for
>
> No.  There will only be a second parse if someone is looking for a filename
> with a trailing dot, and it is not found (*).  I there is no trailing dot in qname, only the first fat_search_long will be called, because len after vfat_striptail_len would be equal to qname->len.

Yes, I'm saying about trailing-dot filename case. Any disadvantage to
standard one is unacceptable to workaround.

>> IO-MEGA, this wouldn't provide proper filename handling.
>
> For accessing IO-MEGA disk in read mode, this is perfect.  I didn't want
> to replicate IO-MEGA write behaviour here, only fix the read behaviour for
> such simple commands as 'ls', 'find' and all the directory browsers.
>
>> If it wants to handle the tailing-dot as a part of filename, it
>> shouldn't be able to access to the stripped-dots filename. (For simple
>> example, I guess you can't do "mv a a." with this patch.)
>
> As I explained above, I only fix read-access on IO-MEGA drives, while
> preserving standard behaviour for write mode.
>
> But I'll try your testcase asap.  Which behaviour do you expect ?
> I would expect a no-op, because I did not change the write-behaviour.

Those sound like strange.  Well, I expect there is no any change to
standard one for IO-MEGA.

And I can't see what is your read-access mean in here. What did this
expect to behave like e.g. following operations,

	$ ls
        a.. a. a
        $ rm -rf *

        $ ls
        a..
        $ touch a.
        $ touch a
...

I assumed you want to define "a." and "a" are different name on
"mv a a.", and _totally_.

Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: [PATCH vfat] allow retrieving entries with trailing dots

[Philippe De Muyter]

Hi Ogawa,

On Thu, Mar 11, 2010 at 02:16:26AM +0900, OGAWA Hirofumi wrote:
> Philippe De Muyter <phdm@macqel.be> writes:
> 
> >> This introduces unneeded directory-parse to standard one.  And for
> >
> > No.  There will only be a second parse if someone is looking for a filename
> > with a trailing dot, and it is not found (*).  I there is no trailing dot in qname, only the first fat_search_long will be called, because len after vfat_striptail_len would be equal to qname->len.
> 
> Yes, I'm saying about trailing-dot filename case. Any disadvantage to
> standard one is unacceptable to workaround.

This is unavoidable.
Let me explain again, with better words I hope,
IOMEGA firmware allows creation of files/directories whose name have
trailing dots.

When I connect such a disk (which is vfat-formatted) via USB to my
linux computer, I want consistency between what getdents says and what
open or stat will do : if getdents says there is a file called "a.",
currently open("a.") and stat("a.") fail with ENOENT No such file or directory.
I would like open("a.") or stat("a.") to succeed.  Nothing more.
For that, I need to first search for the not truncated name.  If that fails
and the filename can be truncated then and only then, try again with the
truncated name.  I cannot let getdents truncate the name, because there 
could also be a different file really called "a".

There would be no advantage in making the change in fat_search_long,
because while searching for "a.", even if it finds a "a" entry, it must
continue till the end of the directory to be sure there is no "a.".

> 
> >> IO-MEGA, this wouldn't provide proper filename handling.
> >
> > For accessing IO-MEGA disk in read mode, this is perfect.  I didn't want
> > to replicate IO-MEGA write behaviour here, only fix the read behaviour for
> > such simple commands as 'ls', 'find' and all the directory browsers.
> >
> >> If it wants to handle the tailing-dot as a part of filename, it
> >> shouldn't be able to access to the stripped-dots filename. (For simple
> >> example, I guess you can't do "mv a a." with this patch.)
> >
> > As I explained above, I only fix read-access on IO-MEGA drives, while
> > preserving standard behaviour for write mode.
> >
> > But I'll try your testcase asap.  Which behaviour do you expect ?
> > I would expect a no-op, because I did not change the write-behaviour.
> 
> Those sound like strange.  Well, I expect there is no any change to
> standard one for IO-MEGA.
> 
> And I can't see what is your read-access mean in here. What did this
> expect to behave like e.g. following operations,

Sorry bad wording. I meant I do not want to change the behaviour for file
creation, only to fix the behaviour while accessi8ng existing files.

> 
> 	$ ls
>         a.. a. a
>         $ rm -rf *
> 
>         $ ls
>         a..
>         $ touch a.
>         $ touch a
> ...
> 
> I assumed you want to define "a." and "a" are different name on
> "mv a a.", and _totally_.

For file creation and renaming, I want to introduce no change, because there
is no problem.  If one wants to create a "a." file on a IO-MEGA disk suing
linux and USB, it is currently called "a", and that will remain exactly the
same.

Best regards

Philippe

Re: [PATCH vfat] allow retrieving entries with trailing dots

[OGAWA Hirofumi]

Philippe De Muyter <phdm@macqel.be> writes:

>> >> This introduces unneeded directory-parse to standard one.  And for
>> >
>> > No.  There will only be a second parse if someone is looking for a filename
>> > with a trailing dot, and it is not found (*).  I there is no trailing dot in qname, only the first fat_search_long will be called, because len after vfat_striptail_len would be equal to qname->len.
>> 
>> Yes, I'm saying about trailing-dot filename case. Any disadvantage to
>> standard one is unacceptable to workaround.
>
> This is unavoidable.

We would be able to introduce new mount option to do it if needed.

>> 	$ ls
>>         a.. a. a
>>         $ rm -rf *
>> 
>>         $ ls
>>         a..
>>         $ touch a.
>>         $ touch a
>> ...
>> 
>> I assumed you want to define "a." and "a" are different name on
>> "mv a a.", and _totally_.
>
> For file creation and renaming, I want to introduce no change, because there
> is no problem.  If one wants to create a "a." file on a IO-MEGA disk suing
> linux and USB, it is currently called "a", and that will remain exactly the
> same.

This changed vfat_find(), so, this patch will change the behavior of all
callers more or less. And the behavior seems to be really strange, you
can remove "a.", but you can't create it? The behavior sounds random,
right?

Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: [PATCH vfat] allow retrieving entries with trailing dots

[Philippe De Muyter]

Hello Ogawa,

On Thu, Mar 11, 2010 at 06:26:54PM +0900, OGAWA Hirofumi wrote:
> Philippe De Muyter <phdm@macqel.be> writes:
> 
> >> >> This introduces unneeded directory-parse to standard one.  And for
> >> >
> >> > No.  There will only be a second parse if someone is looking for a filename
> >> > with a trailing dot, and it is not found (*).  I there is no trailing dot in qname, only the first fat_search_long will be called, because len after vfat_striptail_len would be equal to qname->len.
> >> 
> >> Yes, I'm saying about trailing-dot filename case. Any disadvantage to
> >> standard one is unacceptable to workaround.
> >
> > This is unavoidable.

We could perhaps reduce the disadvantage by moving the logic (test first
with the trailing dots, and then without if needed) into fat_search_long
instead of putting it in vfat_find as my patch proposal does.  This way
the name to find would only be compared once to unrelated entries,
instead of possibly twice as my patch does.

But therefore I need your help : fat_search_long isn't easy to read/modify.

The logic could be modified as such :
    - if the searched name contains trailing dots, first compare
	the truncated part to the same length of the directory entry
    - if they are the same, test the length of the rest of the directory entry
	- if length_of_rest is 0, this could be the matching entry,
	    but there could be a better one later; keep searching till the end.
	- if length_of_rest is not 0, compare the rest with the trailing dots
	    - if rests are equal, we have found it; return
	    - if rests are unequal, continue searching

> 
> We would be able to introduce new mount option to do it if needed.

A new mount option should be the last programming option.  It is better to
automatically do the right thing than to require the user to figure out
he must add a mount option in /etc/fstab or whatever.  Remember this is
for hot-plug disks.

> 
> >> 	$ ls
> >>         a.. a. a
> >>         $ rm -rf *
> >> 
> >>         $ ls
> >>         a..
> >>         $ touch a.
> >>         $ touch a
> >> ...
> >> 
> >> I assumed you want to define "a." and "a" are different name on
> >> "mv a a.", and _totally_.
> >
> > For file creation and renaming, I want to introduce no change, because there
> > is no problem.  If one wants to create a "a." file on a IO-MEGA disk suing
> > linux and USB, it is currently called "a", and that will remain exactly the
> > same.
> 
> This changed vfat_find(), so, this patch will change the behavior of all
> callers more or less. And the behavior seems to be really strange, you
> can remove "a.", but you can't create it?

Yes, you can remove any existing file, but if you want to create a file, 
the name creation rules are kept unchanged. So, creating "a." will
succeed, it will actually be called "a" on disk, but you can still refer
to it as "a." : that will succeed.  That's the strange part of the behavior
but that part is already present for compatibility reasons nor you nor me
can do anything about :(

> The behavior sounds random, right?

It's a compromise to avoid creating name entries that are not universally
accepted, but to allow accessing existing files and directories.

The mount option could be useful then to allow the creation of file and
directory names with trailing dot, but consistency between getdents and
stat/open must be automatic.

Best regards

Philippe

Re: [PATCH vfat] allow retrieving entries with trailing dots

[OGAWA Hirofumi]

Philippe De Muyter <phdm@macqel.be> writes:

>> > This is unavoidable.
>
> We could perhaps reduce the disadvantage by moving the logic (test first
> with the trailing dots, and then without if needed) into fat_search_long
> instead of putting it in vfat_find as my patch proposal does.  This way
> the name to find would only be compared once to unrelated entries,
> instead of possibly twice as my patch does.
>
> But therefore I need your help : fat_search_long isn't easy to read/modify.
>
> The logic could be modified as such :
>     - if the searched name contains trailing dots, first compare
> 	the truncated part to the same length of the directory entry
>     - if they are the same, test the length of the rest of the directory entry
> 	- if length_of_rest is 0, this could be the matching entry,
> 	    but there could be a better one later; keep searching till the end.
> 	- if length_of_rest is not 0, compare the rest with the trailing dots
> 	    - if rests are equal, we have found it; return
> 	    - if rests are unequal, continue searching
>
>> 
>> We would be able to introduce new mount option to do it if needed.
>
> A new mount option should be the last programming option.  It is better to
> automatically do the right thing than to require the user to figure out
> he must add a mount option in /etc/fstab or whatever.  Remember this is
> for hot-plug disks.

Sorry, but I'm not thinking this is primary one. So, requiring option
for it to avoid disadvantage of normal users, it sounds good.

>> This changed vfat_find(), so, this patch will change the behavior of all
>> callers more or less. And the behavior seems to be really strange, you
>> can remove "a.", but you can't create it?
>
> Yes, you can remove any existing file, but if you want to create a file, 
> the name creation rules are kept unchanged. So, creating "a." will
> succeed, it will actually be called "a" on disk, but you can still refer
> to it as "a." : that will succeed.  That's the strange part of the behavior
> but that part is already present for compatibility reasons nor you nor me
> can do anything about :(
>
>> The behavior sounds random, right?
>
> It's a compromise to avoid creating name entries that are not universally
> accepted, but to allow accessing existing files and directories.
>
> The mount option could be useful then to allow the creation of file and
> directory names with trailing dot, but consistency between getdents and
> stat/open must be automatic.

No, this breaks consistency. With this patch, unlink("a."), then
open("a.", O_CREAT) and write(), the result depend on existent
files. This patch is providing two files on one name.

Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: [PATCH vfat] allow retrieving entries with trailing dots

[Philippe De Muyter]

Hello Ogawa,

On Thu, Mar 11, 2010 at 09:41:33PM +0900, OGAWA Hirofumi wrote:
> Philippe De Muyter <phdm@macqel.be> writes:
> 
> >> > This is unavoidable.
> >
> > We could perhaps reduce the disadvantage by moving the logic (test first
> > with the trailing dots, and then without if needed) into fat_search_long
> > instead of putting it in vfat_find as my patch proposal does.  This way
> > the name to find would only be compared once to unrelated entries,
> > instead of possibly twice as my patch does.
> >
> > But therefore I need your help : fat_search_long isn't easy to read/modify.
> >
> > The logic could be modified as such :
> >     - if the searched name contains trailing dots, first compare
> > 	the truncated part to the same length of the directory entry
> >     - if they are the same, test the length of the rest of the directory entry
> > 	- if length_of_rest is 0, this could be the matching entry,
> > 	    but there could be a better one later; keep searching till the end.
> > 	- if length_of_rest is not 0, compare the rest with the trailing dots
> > 	    - if rests are equal, we have found it; return
> > 	    - if rests are unequal, continue searching
> >
> >> 
> >> We would be able to introduce new mount option to do it if needed.
> >
> > A new mount option should be the last programming option.  It is better to
> > automatically do the right thing than to require the user to figure out
> > he must add a mount option in /etc/fstab or whatever.  Remember this is
> > for hot-plug disks.
> 
> Sorry, but I'm not thinking this is primary one. So, requiring option
> for it to avoid disadvantage of normal users, it sounds good.
> 
> >> This changed vfat_find(), so, this patch will change the behavior of all
> >> callers more or less. And the behavior seems to be really strange, you
> >> can remove "a.", but you can't create it?
> >
> > Yes, you can remove any existing file, but if you want to create a file, 
> > the name creation rules are kept unchanged. So, creating "a." will
> > succeed, it will actually be called "a" on disk, but you can still refer
> > to it as "a." : that will succeed.  That's the strange part of the behavior
> > but that part is already present for compatibility reasons nor you nor me
> > can do anything about :(
> >
> >> The behavior sounds random, right?
> >
> > It's a compromise to avoid creating name entries that are not universally
> > accepted, but to allow accessing existing files and directories.
> >
> > The mount option could be useful then to allow the creation of file and
> > directory names with trailing dot, but consistency between getdents and
> > stat/open must be automatic.
> 
> No, this breaks consistency. With this patch, unlink("a."), then
> open("a.", O_CREAT) and write(), the result depend on existent
> files. This patch is providing two files on one name.

To avoid that, we could remember that we have found a filename with a trailing
dot (in that directory or in the whole disk), and if that's the case then
we are allowed to create filenames with trailing dots.

Philippe

Re: [PATCH vfat] allow retrieving entries with trailing dots

[OGAWA Hirofumi]

Philippe De Muyter <phdm@macqel.be> writes:

>> No, this breaks consistency. With this patch, unlink("a."), then
>> open("a.", O_CREAT) and write(), the result depend on existent
>> files. This patch is providing two files on one name.
>
> To avoid that, we could remember that we have found a filename with a trailing
> dot (in that directory or in the whole disk), and if that's the case then
> we are allowed to create filenames with trailing dots.

It sounds like dirty hack, and it would make more complex situations.
And the user can umount/mount (or reboot, etc.) before open().

Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: [PATCH vfat] IOMEGA network drive compatibility

[Philippe De Muyter]

Hello Ogawa,

I changed the subject to make it more explicit.

On Sat, Mar 13, 2010 at 10:06:25PM +0900, OGAWA Hirofumi wrote:
> Philippe De Muyter <phdm@macqel.be> writes:
> 
> >> No, this breaks consistency. With this patch, unlink("a."), then
> >> open("a.", O_CREAT) and write(), the result depend on existent
> >> files. This patch is providing two files on one name.
> >
> > To avoid that, we could remember that we have found a filename with a trailing
> > dot (in that directory or in the whole disk), and if that's the case then
> > we are allowed to create filenames with trailing dots.
> 
> It sounds like dirty hack, and it would make more complex situations.
> And the user can umount/mount (or reboot, etc.) before open().

You can call that a hack, but it certainly is better than the current situation.
IOMEGA network drive firmware allows to create two files in the same directory
differing only by the trailing dots.  Suppose you have in the same directory
a file called "123456789" and a file called "123456789." (same name + a
trailing dot).  When afterwards connected via USB to a linux computer,
ls or any GUI equivalent will show you two files, one called "123456789"
and one called "123456789.".  If you compare those two files, linux/vfat
will tell you that they are equal, because the vfat driver will wrongly
access twice the same "123456789" file and never the "123456789." file.

-rwxr-xr-x 1 phdm root     1 2009-10-15 23:44 123456789
-rwxr-xr-x 1 phdm root     1 2009-10-15 23:44 123456789.

If you then decide to remove the "123456789." file (the one with the
trailing dot), because it is an useless copy of your "123456789" file,
linux/vfat will silently remove the "123456789" file.  Afterwards,
if you're lucky, you'll see ls complain :

	ls: cannot access a.: No such file or directory

, but some GUI interfaces won't say anything.

In the listing, ls will show :

-????????? ? ?    ?        ?                ? 123456789.
(The "123456789" does not appear anymore)

but again some GUI won't even show you the name of the "123456789." file

With my proposed patch, at least if getdents tells a user program that
there is a "123456789." entry, further references by open and friends
will access precisely that entry, not another one.  That at least needs to
be fixed, without any option that the user should give to mount.

Best regards

Philippe

Re: [PATCH vfat] IOMEGA network drive compatibility

[OGAWA Hirofumi]

Philippe De Muyter <phdm@macqel.be> writes:

> You can call that a hack, but it certainly is better than the current situation.
> IOMEGA network drive firmware allows to create two files in the same directory
> differing only by the trailing dots.  Suppose you have in the same directory
> a file called "123456789" and a file called "123456789." (same name + a
> trailing dot).  When afterwards connected via USB to a linux computer,
> ls or any GUI equivalent will show you two files, one called "123456789"
> and one called "123456789.".  If you compare those two files, linux/vfat
> will tell you that they are equal, because the vfat driver will wrongly
> access twice the same "123456789" file and never the "123456789." file.
>
> -rwxr-xr-x 1 phdm root     1 2009-10-15 23:44 123456789
> -rwxr-xr-x 1 phdm root     1 2009-10-15 23:44 123456789.
>
> If you then decide to remove the "123456789." file (the one with the
> trailing dot), because it is an useless copy of your "123456789" file,
> linux/vfat will silently remove the "123456789" file.  Afterwards,
> if you're lucky, you'll see ls complain :
>
> 	ls: cannot access a.: No such file or directory
>
> , but some GUI interfaces won't say anything.
>
> In the listing, ls will show :
>
> -????????? ? ?    ?        ?                ? 123456789.
> (The "123456789" does not appear anymore)
>
> but again some GUI won't even show you the name of the "123456789." file

Because it's wrong entry as fatfs. In this situation, user needs to
repair by tools (e.g. fsck), so it's not responsibility of fs driver.

> With my proposed patch, at least if getdents tells a user program that
> there is a "123456789." entry, further references by open and friends
> will access precisely that entry, not another one.  That at least needs to
> be fixed, without any option that the user should give to mount.

Now, you try to change that design. So, you need to change/think more
high level consistency/design, not just workaround. Otherwise, it'll be
unmaintainable/unfixable.

Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: [PATCH vfat] IOMEGA network drive compatibility

[Philippe De Muyter]

On Sun, Mar 14, 2010 at 08:17:30PM +0900, OGAWA Hirofumi wrote:
> Philippe De Muyter <phdm@macqel.be> writes:
> 
> > You can call that a hack, but it certainly is better than the current situation.
> > IOMEGA network drive firmware allows to create two files in the same directory
> > differing only by the trailing dots.  Suppose you have in the same directory
> > a file called "123456789" and a file called "123456789." (same name + a
> > trailing dot).  When afterwards connected via USB to a linux computer,
> > ls or any GUI equivalent will show you two files, one called "123456789"
> > and one called "123456789.".  If you compare those two files, linux/vfat
> > will tell you that they are equal, because the vfat driver will wrongly
> > access twice the same "123456789" file and never the "123456789." file.
> >
> > -rwxr-xr-x 1 phdm root     1 2009-10-15 23:44 123456789
> > -rwxr-xr-x 1 phdm root     1 2009-10-15 23:44 123456789.
> >
> > If you then decide to remove the "123456789." file (the one with the
> > trailing dot), because it is an useless copy of your "123456789" file,
> > linux/vfat will silently remove the "123456789" file.  Afterwards,
> > if you're lucky, you'll see ls complain :
> >
> > 	ls: cannot access a.: No such file or directory
wrong editing, of course it was :
	ls: cannot access 123456789.: No such file or directory
> >
> > , but some GUI interfaces won't say anything.
> >
> > In the listing, ls will show :
> >
> > -????????? ? ?    ?        ?                ? 123456789.
> > (The "123456789" does not appear anymore)
> >
> > but again some GUI won't even show you the name of the "123456789." file
> 
> Because it's wrong entry as fatfs. In this situation, user needs to
> repair by tools (e.g. fsck), so it's not responsibility of fs driver.

I know that entries with trailing dots are "invalid" fatfs entries,
but 1) linux/vfat currently accept such entries silently without truncating
them and 2) IOMEGA disks present themselves as fatfs disks.  Who are we to
tell the opposite ?  And IOMEGA integrated scandisk (accessible via the
web interface) does not find any error in its filesystem.

> 
> > With my proposed patch, at least if getdents tells a user program that
> > there is a "123456789." entry, further references by open and friends
> > will access precisely that entry, not another one.  That at least needs to
> > be fixed, without any option that the user should give to mount.
> 
> Now, you try to change that design. So, you need to change/think more
> high level consistency/design, not just workaround. Otherwise, it'll be
> unmaintainable/unfixable.

I am sorry but I don't see anything unmaintainable/unfixable in my patch :
it is very small and localized and it only allows accessing existing
entries with trailing dots, nothing more.
It does not allow to create "invalid" entries, so it does not change the
behaviour for "strict" vfat disks, it only introduces a small time penalty
when a user tries to access a file using a different name that the one stored
in the directory, which happens very rarely : most linux users use GUI's or
bash tab-completion to access files, which implies they use the name given by
the file-system driver, not a variation of it.

Now if we are concerned about "editing" such files by creating a temporary
file and renaming, we can look at what is needed to make renaming to an
existing name containing trailing dots succeed.  I have not tested that yet.

For the rest, i.e. creating arbitrary names with trailing dots just like
what IOMEGA drives do, it can be with a mount option, or left as it is
now : impossible.

Philippe

Re: [PATCH vfat] IOMEGA network drive compatibility

[OGAWA Hirofumi]

Philippe De Muyter <phdm@macqel.be> writes:

>> Because it's wrong entry as fatfs. In this situation, user needs to
>> repair by tools (e.g. fsck), so it's not responsibility of fs driver.
>
> I know that entries with trailing dots are "invalid" fatfs entries,
> but 1) linux/vfat currently accept such entries silently without truncating
> them and 2) IOMEGA disks present themselves as fatfs disks.  Who are we to
> tell the opposite ?  And IOMEGA integrated scandisk (accessible via the
> web interface) does not find any error in its filesystem.

You can create many files points to one cluster chain. It is silently
accepted by current fatfs driver. But, it's a broken fs image like this.

>> Now, you try to change that design. So, you need to change/think more
>> high level consistency/design, not just workaround. Otherwise, it'll be
>> unmaintainable/unfixable.
>
> I am sorry but I don't see anything unmaintainable/unfixable in my patch :
> it is very small and localized and it only allows accessing existing
> entries with trailing dots, nothing more.
>
> It does not allow to create "invalid" entries, so it does not change the
> behaviour for "strict" vfat disks, it only introduces a small time penalty
> when a user tries to access a file using a different name that the one stored
> in the directory, which happens very rarely : most linux users use GUI's or
> bash tab-completion to access files, which implies they use the name given by
> the file-system driver, not a variation of it.
>
> Now if we are concerned about "editing" such files by creating a temporary
> file and renaming, we can look at what is needed to make renaming to an
> existing name containing trailing dots succeed.  I have not tested that yet.
>
> For the rest, i.e. creating arbitrary names with trailing dots just like
> what IOMEGA drives do, it can be with a mount option, or left as it is
> now : impossible.

I already reported the one of bugs at least, and it seems you already
can't fix it, or just ignored it.

You are always saying "I'm fine with this". But I disagree, and I know
this fs is not only for just desktop users.  If this is only for you
usage, I'm sorry, but please apply it yourself.

Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>