some doubts in dokanfuse

[FaerieLee]

Feature request can skip this form. Bug report must complete it. Check List must be 100% match or it will be automatically closed without further discussion. Please remove this line.

Environment

• Windows version: Win11
• Processor architecture: x64
• Dokany version: any
• Library type (Dokany/FUSE): FUSE

Check List

• [ Y] I checked my issue doesn't exist yet
• [ Y] My issue is valid with mirror default sample and not specific to my user-mode driver implementation
• [ Y] I can always reproduce the issue with the provided description below.
• [ Y] I have updated Dokany to the latest version and have reboot my computer after.
• [ Y] I tested one of the last snapshot from appveyor CI

Description

Question1: The following is the implementation of the function fuse_parse_cmdline

if (mountpoint)
    *mountpoint = hopts.mountpoint;
else
    free(hopts.mountpoint);
if (multithreaded)
    *multithreaded = !hopts.singlethread;
if (foreground)
    *foreground = hopts.foreground;
return 0;

if mountpoint == nullptr, the code wuld free the hots.mountpoint and return 0. In the implementation of funtion fuse_setup

  res = fuse_parse_cmdline(&args, mountpoint, multithreaded, &foreground);
  if (res == -1)
    return nullptr;

  ch = fuse_mount(*mountpoint, &args);

So in the case of that mountpoint == nullptr, it will continue to execute function fuse_mount. the process would crash because of dereferencing null pointer. The dokanfuse.dll doesn't only recognize invalid argument, but also result to crash. I think the code is not robust. what do you think ?

Question2: In the struct fuse_operations, the parameters of init funtion pointer hasn't variable private_data. void *(*init) (struct fuse_conn_info *conn); In the case of using c++ to build filesystem, it's possible to call fuse_setup in the oridinary funtion. but call init only in the static function, how to setup variable private_data which is the member of c++ class in the static function without any information ? I think this design is unreasonable . In the library libfuse, the defination of init function pointer is following: void (*init) (void *userdata, struct fuse_conn_info *conn); So, is it necessary to upgrade init funtion pointer ?

Thanks

Logs

Only some doubts, no log.

[FaerieLee]

Hello，If you think it makes sense, please reply, if not, please reply, looking forward to your reply，Thanks.

[Liryna]

Hi @FaerieLee ,

Sorry I was out when this was created.

1 - Actually the only usage of fuse_parse_cmdline is in fuse_setup->fuse_main_real which always provide a valid variable. https://github.com/dokan-dev/dokany/blob/ce8b8229ad8262f920ca8d284448c1fe10e46cae/dokan_fuse/src/dokanfuse.cpp#L810-L815 The code is "robust" but just does not makes sense now. It is probably a legacy of a previous logic.

2 - I understand but that's how fuse is designed even in the latest API version http://libfuse.github.io/doxygen/structfuse__operations.html#ae2d73c7b5604a9fddffec229001d4556

[FaerieLee]

1 - you are right. but in our code, we direcly use fuse_setup, not fuse_main_real, accroding to the routine of libfuse. So if directly call fuse_setup in the case of mountPoint=nullptr, the process would crash. Is need to add logic to check mountpoint in exported funcion fuse_setup ?

2 - void (init) (void userdata, struct fuse_conn_info conn); belong to the lowlevel api in the libfuse. void ( fuse_operations::init) (struct fuse_conn_info conn, struct fuse_config *cfg) belong to high level api, in the function, it doesn't carry private_data, So current design of dokany is consistent with the high-level api of libfuse. But I still have doubts about the design of libfuse.

[Liryna]

1 - Oh I missed that! Then yeah I believe in fuse_parse_cmdline we should have:

    if (mountpoint)
        *mountpoint = hopts.mountpoint;
    else
+       goto err;

    if (multithreaded)
        *multithreaded = !hopts.singlethread;
    if (foreground)
        *foreground = hopts.foreground;
    return 0;

err:
+   if (hopts.mountpoint)
+     free(hopts.mountpoint);
    return -1;
}

What do you think ? If you agree, I believe it is rightful that you open a pull request since you found this but I can also do it

[FaerieLee]

I think that the modification is perfect. Can i submit the code using your above modification ? It's really a pleasure to be able to contribute to open source projects. I really like it.

[Liryna]

Yes go for it!

[Liryna]

@FaerieLee Have you been able to prepare a pull request ?

[Liryna]

https://github.com/dokan-dev/dokany/commit/fbf335d9804d7502c696e39e55a68a48b326b56c