On Wed, Sep 30, 2020 at 01:57:40PM +0200, Greg Kroah-Hartman wrote:
> Kees, and Rafael, I don't know if you saw this proposal from Joe for
> sysfs files, questions below:
I'm a fan. I think the use of sprintf() in sysfs might have been one of
my earliest complaints about unsafe code patterns in the kernel. ;)
> > +/**
> > + * sysfs_emit - scnprintf equivalent, aware of PAGE_SIZE buffer.
> > + * @buf: start of PAGE_SIZE buffer.
> > + * @fmt: format
> > + * @...: optional arguments to @format
> > + *
> > + *
> > + * Returns number of characters written to @buf.
> > + */
> > +int sysfs_emit(char *buf, const char *fmt, ...)
> > +{
> > + va_list args;
> > + int len;
> > +
> > + if (WARN(!buf || offset_in_page(buf),
> > + "invalid sysfs_emit: buf:%p\n", buf))
I don't want the %p here, but otherwise, sure. I'd also make it a _ONCE
variant:
if (WARN_ONCE(!buf || offset_in_page(buf),
"invalid sysfs_emit: offset_in_page(buf):%zd\n",
buf ? offset_in_page(buf) : 0))
> > + return 0;
> > +
> > + va_start(args, fmt);
> > + len = vscnprintf(buf, PAGE_SIZE, fmt, args);
> > + va_end(args);
> > +
> > + return len;
> > +}
> > +EXPORT_SYMBOL_GPL(sysfs_emit);
> > +
> > +/**
> > + * sysfs_emit_at - scnprintf equivalent, aware of PAGE_SIZE buffer.
> > + * @buf: start of PAGE_SIZE buffer.
> > + * @at: offset in @buf to start write in bytes
> > + * @at must be >= 0 && < PAGE_SIZE
> > + * @fmt: format
> > + * @...: optional arguments to @fmt
> > + *
> > + *
> > + * Returns number of characters written starting at &@buf[@at].
> > + */
> > +int sysfs_emit_at(char *buf, int at, const char *fmt, ...)
> > +{
> > + va_list args;
> > + int len;
> > +
> > + if (WARN(!buf || offset_in_page(buf) || at < 0 || at >= PAGE_SIZE,
> > + "invalid sysfs_emit_at: buf:%p at:%d\n", buf, at))
Same:
if (WARN_ONCE(!buf || offset_in_page(buf) || at < 0 || at >= PAGE_SIZE,
"invalid sysfs_emit_at: offset_in_page(buf):%zd at:%d\n",
buf ? offset_in_page(buf) : 0, at))
> > + return 0;
> > +
> > + va_start(args, fmt);
> > + len = vscnprintf(buf + at, PAGE_SIZE - at, fmt, args);
> > + va_end(args);
> > +
> > + return len;
> > +}
> > +EXPORT_SYMBOL_GPL(sysfs_emit_at);
>
> These feel sane, but I'm loath to have a ton of churn for no good
> reason.
I think the churn is worth it if we remove "seemingly wrong" code
patterns from the kernel. It's especially useful, IMO, for when there
are future mutations/refactorings and we don't end up with a bare
sprintf somewhere else.
> If we make all sysfs show/store functions use these calls instead of
> sprintf(), it "feels" like that might address the objections that people
> have had in the past where they are nervous about "bare" sprintf()
> calls, right?
I would think so. This is the kind of thing we did for %n in seq_file:
remove potential foot-gun API in favor of subsystem-specific safe API.
> It also might make things easier to audit where we can see much easier
> where sysfs files are doing "foolish" things by calling sysfs_emit_at()
> a bunch of times they shouldn't be, and maybe automate the documentation
> of sysfs files in a better way.
Indeed!
> So I guess I'm asking for another developer to at least agree that this
> feels like the right way forward here. I don't want to start down this
> path, only to roll them all back as it feels like pointless churn.
With the changes above, I'd Ack it. :)
--
Kees Cook
