Re: Fw: Re: oops in choose_configuration()

[Greg KH <greg@kroah.com>]

On Wed, Mar 08, 2006 at 01:06:59AM -0500, Dmitry Torokhov wrote:
> On Wednesday 08 March 2006 00:23, Greg KH wrote:
> > > 
> > > Hmm, what is the policy for attr->show()? With hotplug variables we
> > > return -ENOMEM if there is not enough memory to store all data, but
> > > what about attributes? Should we also return error (and which one,
> > > -ENOMEM, -ENOBUFS?) or fill as much as we can and return up to
> > > PAGE_SIZE?
> > 
> > Remember, sysfs files are supposed to be small, you are an "oddity" in
> > that you have a much larger buffer that you can return due to the wierd
> > aliases you have.
> > 
> > Truncating the buffer is probably good as we want userspace to get some
> > information, right?
> > 
> > > With sysfs not kernel nor application can really recover
> > > if attribute needs buffer larger than a page. Or just rely on BUG_ON
> > > in fs/sysfs/file.c::fill_read_buffer()?
> > 
> > How about just making this a binary attribute, then you can handle an
> > arbitrary size buffer and don't have to worry about the PAGE_SIZE stuff
> > (but it makes it more code that you have to write to handle it all,
> > there are tradeoffs...)
> > 
> 
> I really don't believe that we ever going to cross 4096 boundary for any
> single input attribute, but just to code defensively we need to decide
> what to do if we ever encounter a crazy device. Just truncating to
> PAGE_SIZE is easiest so that's what I am going to do.

Ok, that's fine with me.

thanks,

greg k-h