[PATCH] mm/gup.c: Handling ERR within unpin_user_pages()

[PATCH] mm/gup.c: Handling ERR within unpin_user_pages()

[Souptick Joarder]

It is possible that a buggy caller of unpin_user_pages()
(specially in error handling path) may end up calling it with
npages < 0 which is unnecessary.

This can be fixed by adding extra check inside unpin_user_pages().

Signed-off-by: Souptick Joarder <jrdr.linux@gmail.com>
Cc: John Hubbard <jhubbard@nvidia.com>
---
 mm/gup.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/mm/gup.c b/mm/gup.c
index 0b5c308b..2e19bd6 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -328,6 +328,9 @@ void unpin_user_pages(struct page **pages, unsigned long npages)
 {
 	unsigned long index;
 
+	if (WARN_ON_ONCE(npages < 0))
+		return;
+
 	/*
 	 * TODO: this can be optimized for huge pages: if a series of pages is
 	 * physically contiguous and part of the same compound page, then a
-- 
1.9.1

Re: [PATCH] mm/gup.c: Handling ERR within unpin_user_pages()

[Matthew Wilcox]

On Sun, Sep 13, 2020 at 08:02:35PM +0530, Souptick Joarder wrote:
> It is possible that a buggy caller of unpin_user_pages()
> (specially in error handling path) may end up calling it with
> npages < 0 which is unnecessary.
> @@ -328,6 +328,9 @@ void unpin_user_pages(struct page **pages, unsigned long npages)
>  {
>  	unsigned long index;
>  
> +	if (WARN_ON_ONCE(npages < 0))
> +		return;

But npages is unsigned long.  So it can't be less than zero.

Re: [PATCH] mm/gup.c: Handling ERR within unpin_user_pages()

[Souptick Joarder]

On Sun, Sep 13, 2020 at 8:25 PM Matthew Wilcox <willy@infradead.org> wrote:
>
> On Sun, Sep 13, 2020 at 08:02:35PM +0530, Souptick Joarder wrote:
> > It is possible that a buggy caller of unpin_user_pages()
> > (specially in error handling path) may end up calling it with
> > npages < 0 which is unnecessary.
> > @@ -328,6 +328,9 @@ void unpin_user_pages(struct page **pages, unsigned long npages)
> >  {
> >       unsigned long index;
> >
> > +     if (WARN_ON_ONCE(npages < 0))
> > +             return;
>
> But npages is unsigned long.  So it can't be less than zero.

Sorry, I missed it.

Then, it means if npages is assigned with -ERRNO by caller, unpin_user_pages()
may end up calling a big loop, which is unnecessary.

Re: [PATCH] mm/gup.c: Handling ERR within unpin_user_pages()

[Jason Gunthorpe]

On Mon, Sep 14, 2020 at 07:20:34AM +0530, Souptick Joarder wrote:
> On Sun, Sep 13, 2020 at 8:25 PM Matthew Wilcox <willy@infradead.org> wrote:
> >
> > On Sun, Sep 13, 2020 at 08:02:35PM +0530, Souptick Joarder wrote:
> > > It is possible that a buggy caller of unpin_user_pages()
> > > (specially in error handling path) may end up calling it with
> > > npages < 0 which is unnecessary.
> > > @@ -328,6 +328,9 @@ void unpin_user_pages(struct page **pages, unsigned long npages)
> > >  {
> > >       unsigned long index;
> > >
> > > +     if (WARN_ON_ONCE(npages < 0))
> > > +             return;
> >
> > But npages is unsigned long.  So it can't be less than zero.
> 
> Sorry, I missed it.
> 
> Then, it means if npages is assigned with -ERRNO by caller, unpin_user_pages()
> may end up calling a big loop, which is unnecessary.

How will a caller allocate memory of the right size and still manage
to call with the wrong npages? Do you have an example of a broken caller?

Jason

Re: [PATCH] mm/gup.c: Handling ERR within unpin_user_pages()

[Souptick Joarder]

On Mon, Sep 14, 2020 at 7:38 PM Jason Gunthorpe <jgg@ziepe.ca> wrote:
>
> On Mon, Sep 14, 2020 at 07:20:34AM +0530, Souptick Joarder wrote:
> > On Sun, Sep 13, 2020 at 8:25 PM Matthew Wilcox <willy@infradead.org> wrote:
> > >
> > > On Sun, Sep 13, 2020 at 08:02:35PM +0530, Souptick Joarder wrote:
> > > > It is possible that a buggy caller of unpin_user_pages()
> > > > (specially in error handling path) may end up calling it with
> > > > npages < 0 which is unnecessary.
> > > > @@ -328,6 +328,9 @@ void unpin_user_pages(struct page **pages, unsigned long npages)
> > > >  {
> > > >       unsigned long index;
> > > >
> > > > +     if (WARN_ON_ONCE(npages < 0))
> > > > +             return;
> > >
> > > But npages is unsigned long.  So it can't be less than zero.
> >
> > Sorry, I missed it.
> >
> > Then, it means if npages is assigned with -ERRNO by caller, unpin_user_pages()
> > may end up calling a big loop, which is unnecessary.
>
> How will a caller allocate memory of the right size and still manage
> to call with the wrong npages? Do you have an example of a broken caller?

These are two broken callers which might end up calling unpin_user_pages()
with -ERRNO.
drivers/rapidio/devices/rio_mport_cdev.c#L952
drivers/misc/mic/scif/scif_rma.c#L1399

They both are in error handling paths, so might not have any serious impact.
But theoretically possible.

Re: [PATCH] mm/gup.c: Handling ERR within unpin_user_pages()

[John Hubbard]

On 9/14/20 1:52 PM, Souptick Joarder wrote:
> On Mon, Sep 14, 2020 at 7:38 PM Jason Gunthorpe <jgg@ziepe.ca> wrote:
>>
>> On Mon, Sep 14, 2020 at 07:20:34AM +0530, Souptick Joarder wrote:
>>> On Sun, Sep 13, 2020 at 8:25 PM Matthew Wilcox <willy@infradead.org> wrote:
>>>>
>>>> On Sun, Sep 13, 2020 at 08:02:35PM +0530, Souptick Joarder wrote:
>>>>> It is possible that a buggy caller of unpin_user_pages()
>>>>> (specially in error handling path) may end up calling it with
>>>>> npages < 0 which is unnecessary.
>>>>> @@ -328,6 +328,9 @@ void unpin_user_pages(struct page **pages, unsigned long npages)
>>>>>   {
>>>>>        unsigned long index;
>>>>>
>>>>> +     if (WARN_ON_ONCE(npages < 0))
>>>>> +             return;
>>>>
>>>> But npages is unsigned long.  So it can't be less than zero.
>>>
>>> Sorry, I missed it.
>>>
>>> Then, it means if npages is assigned with -ERRNO by caller, unpin_user_pages()
>>> may end up calling a big loop, which is unnecessary.
>>
>> How will a caller allocate memory of the right size and still manage
>> to call with the wrong npages? Do you have an example of a broken caller?
> 
> These are two broken callers which might end up calling unpin_user_pages()
> with -ERRNO.
> drivers/rapidio/devices/rio_mport_cdev.c#L952
> drivers/misc/mic/scif/scif_rma.c#L1399
> 
> They both are in error handling paths, so might not have any serious impact.
> But theoretically possible.
> 

Eventually, I settled on fixing up the callers so that they match the gup/pup
API better. In other words, gup/pup has signed int for both input and return
value, and the callers need to handle that perfectly.

So let's fix up the callers.

thanks,
-- 
John Hubbard
NVIDIA

Re: [PATCH] mm/gup.c: Handling ERR within unpin_user_pages()

[Ira Weiny]

On Tue, Sep 15, 2020 at 02:22:33AM +0530, Souptick Joarder wrote:
> On Mon, Sep 14, 2020 at 7:38 PM Jason Gunthorpe <jgg@ziepe.ca> wrote:
> >
> > On Mon, Sep 14, 2020 at 07:20:34AM +0530, Souptick Joarder wrote:
> > > On Sun, Sep 13, 2020 at 8:25 PM Matthew Wilcox <willy@infradead.org> wrote:
> > > >
> > > > On Sun, Sep 13, 2020 at 08:02:35PM +0530, Souptick Joarder wrote:
> > > > > It is possible that a buggy caller of unpin_user_pages()
> > > > > (specially in error handling path) may end up calling it with
> > > > > npages < 0 which is unnecessary.
> > > > > @@ -328,6 +328,9 @@ void unpin_user_pages(struct page **pages, unsigned long npages)
> > > > >  {
> > > > >       unsigned long index;
> > > > >
> > > > > +     if (WARN_ON_ONCE(npages < 0))
> > > > > +             return;
> > > >
> > > > But npages is unsigned long.  So it can't be less than zero.
> > >
> > > Sorry, I missed it.
> > >
> > > Then, it means if npages is assigned with -ERRNO by caller, unpin_user_pages()
> > > may end up calling a big loop, which is unnecessary.
> >
> > How will a caller allocate memory of the right size and still manage
> > to call with the wrong npages? Do you have an example of a broken caller?
> 
> These are two broken callers which might end up calling unpin_user_pages()
> with -ERRNO.
> drivers/rapidio/devices/rio_mport_cdev.c#L952

The error here is that nr_pages should not be set to pinned if pinned is < 0.

Why not fix the logic there?  Because it is inherently dangerous to set an
unsigned from a signed variable like that.

> drivers/misc/mic/scif/scif_rma.c#L1399

Again this is a caller who is not properly checking error conditions.

> 
> They both are in error handling paths, so might not have any serious impact.
> But theoretically possible.

Actually I think they might have serious problems so they both should be fixed.

In the end this patch just can't work because npages can't be < 0 like Matthew
said.

Ira