[PATCH] nvme: fix string_matches() helper

Christoph Hellwig hch at lst.de
Tue Jun 13 23:38:23 PDT 2017


On Tue, Jun 13, 2017 at 08:42:59AM -0700, Andy Lutomirski wrote:
> On Tue, Jun 13, 2017 at 6:07 AM, Christoph Hellwig <hch at lst.de> wrote:
> > NVMe "ASCII" strings are not nul-terminated and can use up every single
> > byte in the field.  Thus use strnlen to determine the match length instead
> > of possibly overrunning the field.
> >
> > Signed-off-by: Christoph Hellwig <hch at lst.de>
> > ---
> >  drivers/nvme/host/core.c | 4 +---
> >  1 file changed, 1 insertion(+), 3 deletions(-)
> >
> > diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
> > index 434b18863895..2658a3a3cbb5 100644
> > --- a/drivers/nvme/host/core.c
> > +++ b/drivers/nvme/host/core.c
> > @@ -1581,9 +1581,7 @@ static bool string_matches(const char *idstr, const char *match, size_t len)
> >         if (!match)
> >                 return true;
> >
> > -       matchlen = strlen(match);
> > -       WARN_ON_ONCE(matchlen > len);
> > -
> > +       matchlen = strnlen(match, len);
> 
> "match" refers to the string in the quirk table, which should be a
> plain C string.  Are you hitting this in practice?

No, but I've just written some other code that deals with NVMe
strings, and took extra care of bounds protection.

But yes, given the strlen is on the match we should be ok as long
as the WARN_ON_ONCE also does an early return.  Does that sound ok
to you?



More information about the Linux-nvme mailing list