Re: [GIT] /proc/uptime fix

[Alexey Dobriyan <adobriyan@gmail.com>]

On Mon, Nov 03, 2008 at 10:14:16AM -0800, Linus Torvalds wrote:
> On Mon, 3 Nov 2008, Alexey Dobriyan wrote:
> > 
> > Alexey Dobriyan (1):
> >       proc: revert /proc/uptime to ->read_proc hook
> 
> Ok, pulled, but it's a bit sad. 
> 
> Al: the commit is 6c87df37dcb9c6c33923707fa5191e0a65874d60, and the 
> message explains it all:
> 
>     Turned out some VMware userspace does pread(2) on /proc/uptime, but
>     seqfiles currently don't allow pread() resulting in -ESPIPE.
>     
>     Seqfiles in theory can do pread(), but this can be a long story,
>     so revert to ->read_proc until then.
>     
> and I bet that the pread() usage is just with a constant loff_t of 0, and 
> just there to avoid a simple case of "lseek+read". 
> 
> I wonder how hard it is to make seqfiles support pread.

Not hard at all, if ->read hook has information if it's called from read(2)
or pread(2) (or something equivalent).

> Maybe it's something as trivial as just flushing the buffer whenever 'pos'
> doesn't match f_pos?

> --- a/fs/seq_file.c
> +++ b/fs/seq_file.c
> @@ -48,8 +48,8 @@ int seq_open(struct file *file, const struct seq_operations *op)
>  	 */
>  	file->f_version = 0;
>  
> -	/* SEQ files support lseek, but not pread/pwrite */
> -	file->f_mode &= ~(FMODE_PREAD | FMODE_PWRITE);
> +	/* SEQ files support lseek, but not pwrite */
> +	file->f_mode &= ~(FMODE_PWRITE);

FMODE_PWRITE is aliased to FMODE_PREAD so you have to split them first. ;-)

> @@ -91,6 +91,8 @@ ssize_t seq_read(struct file *file, char __user *buf, size_t size, loff_t *ppos)
>  		if (!m->buf)
>  			goto Enomem;
>  	}
> +	if (pos != file->f_pos)
> +		m->count = 0;
>  	/* if not empty - flush it first */
>  	if (m->count) {
>  		n = min(m->count, size);
> @@ -175,6 +177,8 @@ Done:
>  	else
>  		*ppos += copied;
>  	file->f_version = m->version;
> +	if (pos != file->f_pos)
> +		m->count = 0;