Re: [fuse-devel] open flags are not passed correctly to CUSE open callback functio

[Stefan S. <dev...@gm...>]

Hi,

On 07/09/2012 02:20 PM, Goswin von Brederlow wrote:
> On Fri, Jul 06, 2012 at 08:18:04PM +0200, Stefan Schoenleitner wrote:
>> Hi,
>>
>> according to the documentation [1,2] the CUSE/FUSE open callback
>> function includes the flags that were given to the open() call.
>> To my surprise these flags do not seem to be passed correctly.
>> I used strace to trace a simple 'echo test > /dev/cusedev' and discovered that
>> the flags provided to the open call are O_WRONLY|O_CREAT|O_TRUNC (0x0241).
>> However, the flags in the CUSE open callback (fi->flags) are set to 0x8001
>> which leads to wrong results of course.
>> Is this a bug in fuse-2.9.0, or is this intentional for some reason ?
>>
>> Cheers,
>> Stefan
>>
>> [1] http://sourceforge.net/apps/mediawiki/fuse/index.php?title=Fuse_file_info
>> [2] http://fuse.sourceforge.net/doxygen/structfuse__file__info.html
> 
> /usr/include/fuse/fuse.h
> 
>         /** File open operation
>          *
>          * No creation (O_CREAT, O_EXCL) and by default also no
>          * truncation (O_TRUNC) flags will be passed to open(). If an
>          * application specifies O_TRUNC, fuse first calls truncate()
>          * and then open(). Only if 'atomic_o_trunc' has been
>          * specified and kernel version is 2.6.24 or later, O_TRUNC is
>          * passed on to open.

thank you for the information.
For CUSE, there is no truncate() or setattr() callback.
How do I know whether these flags were specified in the open() call to
the CUSE device ?
For testing purposes, I forwarded any open/read/write calls to a regular
file.
In case truncate was called on the file, it seems to have had no effect.
In case of a write operation to the CUSE device, the file is overwritten
with the content, but not truncated.


In my proxy, I guess I could just always use these flags whenever they
make sense.
So, if there is a call to open, I would always specify O_CREAT in the
call to the real open().
And I would also add O_TRUNC, in case the file exists and O_APPEND is
not set.
Do you think this would work ?

However, I wonder if truncation or file creation is relevant at all for
character devices.
After all, my goal is to pass on any communication with a character
device to the (existing) character device on a remote host.
Maybe this is also the reason why these flags are not passed on by
FUSE/CUSE ?


> 
> MfG
> 	Goswin


Cheers,
Stefan