Re: [RFC PATCH 2/3] add statmnt(2) syscall

From: Miklos Szeredi
Date: Tue Sep 19 2023 - 09:00:13 EST


On Tue, 19 Sept 2023 at 14:41, Christian Brauner <brauner@xxxxxxxxxx> wrote:
>
> > > with __u32 size for mnt_root and mnt_point
> >
> > Unnecessary if the strings are nul terminated.
>
> All ok by me so far but how does the kernel know the size of the buffer
> to copy into? Wouldn't it be better to allow userspace to specify that?
> I'm probably just missing something but I better ask.

Because size of the buffer is given as the syscall argument.

long statmount(u64 mnt_id, u64 mask, struct statmnt __user *buf,
size_t bufsize, unsigned int flags);

If you are still hung up about this not being properly typed, how about this:

struct statmnt {
__u64 mask; /* What results were written [uncond] */
__u32 sb_dev_major; /* Device ID */
[...]
__u64 fs_type; /* [str] Filesystem type */
__u64 __spare[49];
char __string_buf[];
};

Such variable length structures are used all over the place, this
isn't some big invention. The only new thing is that we set pointers
to within the tail part of the buffer, to make the interface work for
the multiple strings case.

Thanks,
Miklos