Re: [Qemu-devel] [PATCH] linux-user: add name_to_handle_at/open_by_handle_at

[Peter Maydell]

On 27 August 2015 at 23:27, Laurent Vivier <address@hidden> wrote:
> This patch allows to run example given by open_by_handle_at(2):
>
>       The following shell session demonstrates the use of these two programs:
>
>            $ echo 'Can you please think about it?' > cecilia.txt
>            $ ./t_name_to_handle_at cecilia.txt > fh
>            $ ./t_open_by_handle_at < fh
>            open_by_handle_at: Operation not permitted
>            $ sudo ./t_open_by_handle_at < fh      # Need CAP_SYS_ADMIN
>            Read 31 bytes
>            $ rm cecilia.txt
>
>        Now  we delete and (quickly) re-create the file so that it has the same
>        content and (by chance) the  same  inode.[...]
>
>            $ stat --printf="%i\n" cecilia.txt     # Display inode number
>            4072121
>            $ rm cecilia.txt
>            $ echo 'Can you please think about it?' > cecilia.txt
>            $ stat --printf="%i\n" cecilia.txt     # Check inode number
>            4072121
>            $ sudo ./t_open_by_handle_at < fh
>            open_by_handle_at: Stale NFS file handle
>
> See the man page for source code.
>
> Signed-off-by: Laurent Vivier <address@hidden>
> ---
>  linux-user/syscall.c | 97 
> ++++++++++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 97 insertions(+)
>
> diff --git a/linux-user/syscall.c b/linux-user/syscall.c
> index f62c698..725ed66 100644
> --- a/linux-user/syscall.c
> +++ b/linux-user/syscall.c
> @@ -5246,6 +5246,93 @@ static int do_futex(target_ulong uaddr, int op, int 
> val, target_ulong timeout,
>          return -TARGET_ENOSYS;
>      }
>  }
> +#if defined(TARGET_NR_name_to_handle_at) && defined(CONFIG_OPEN_BY_HANDLE)
> +static abi_long do_name_to_handle_at(abi_long arg1, abi_long arg2,
> +                                     abi_long arg3, abi_long arg4,
> +                                     abi_long arg5)

Since this is in its own function you have the opportunity to
give the input parameters more meaningful names than arg1..arg5.

> +{
> +    struct file_handle *target_fh;
> +    struct file_handle *fh;
> +    int mount_id = 0;
> +    abi_long ret;
> +    char *name;
> +    unsigned int size;
> +
> +    if (get_user_s32(size, arg3)) {
> +        return -TARGET_EFAULT;
> +    }
> +
> +    name = lock_user_string(arg2);
> +    if (!name) {
> +        return -TARGET_EFAULT;
> +    }
> +
> +    target_fh = lock_user(VERIFY_WRITE, arg3,
> +                          sizeof(struct file_handle) + size, 0);
> +    if (!target_fh) {
> +        unlock_user(name, arg2, 0);
> +        return -TARGET_EFAULT;
> +    }
> +
> +    fh = g_malloc0(sizeof(struct file_handle) + size);
> +    fh->handle_bytes = size;

I was going to suggest just using target_fh, since we know
the host and guest have the same sized struct here. But I
guess they might have different alignment restrictions.

> +
> +    ret = get_errno(name_to_handle_at(arg1, path(name), fh, &mount_id, 
> arg5));
> +    unlock_user(name, arg2, 0);
> +
> +    /* man name_to_handle_at(2):
> +     * Other than the use of the handle_bytes field, the caller should treat
> +     * the file_handle structure as an opaque data type
> +     */
> +
> +    memcpy(target_fh, fh, fh->handle_bytes);
> +    target_fh->handle_bytes = tswap32(fh->handle_bytes);

You need to swap the handle_type field too (to match the
swap you do in do_open_by_handle_at()).
(I think swapping in both places is better than doing so in
neither.)

Otherwise looks good.

thanks
-- PMM