Re: [PATCH v2] seq_file: Unconditionally use vmalloc for buffer

From: Greg Kroah-Hartman
Date: Wed Mar 17 2021 - 11:54:36 EST

On Wed, Mar 17, 2021 at 03:44:16PM +0100, Michal Hocko wrote:
> On Wed 17-03-21 14:34:27, Greg KH wrote:
> > On Wed, Mar 17, 2021 at 01:08:21PM +0100, Michal Hocko wrote:
> > > Btw. I still have problems with the approach. seq_file is intended to
> > > provide safe way to dump values to the userspace. Sacrificing
> > > performance just because of some abuser seems like a wrong way to go as
> > > Al pointed out earlier. Can we simply stop the abuse and disallow to
> > > manipulate the buffer directly? I do realize this might be more tricky
> > > for reasons mentioned in other emails but this is definitely worth
> > > doing.
> >
> > We have to provide a buffer to "write into" somehow, so what is the best
> > way to stop "abuse" like this?
> What is wrong about using seq_* interface directly?

Right now every show() callback of sysfs would have to be changed :(

> > Right now, we do have helper functions, sysfs_emit(), that know to stop
> > the overflow of the buffer size, but porting the whole kernel to them is
> > going to take a bunch of churn, for almost no real benefit except a
> > potential random driver that might be doing bad things here that we have
> > not noticed yet.
> I am not familiar with sysfs, I just got lost in all the indirection but
> replacing buffer by the seq_file and operate on that should be possible,
> no?

sysfs files should be very simple and easy, and have a single value
being written to userspace. I guess seq_printf() does handle the issue
of "big buffers", but there should not be a big buffer here to worry
about in the first place (yes, there was a bug where a driver took
unchecked data and sent it to userspace overflowing the buffer which
started this whole thread...)

I guess Kees wants to change all show functions to use the seq_ api,
which now makes a bit more sense, but still seems like a huge overkill.
But I now understand the idea here, the buffer management is handled by
the core kernel and overflows are impossible.

A "simpler" fix is to keep the api the same today, and just "force"
everyone to use sysfs_emit() which does the length checking

I don't know, it all depends on how much effort we want to put into the
"drivers can not do stupid things because we prevent them from it"
type of work here...


greg k-h