[PATCH 1/1] Fix: vmw_vmci driver get_user_pages_fast error handling

[PATCH 1/1] Fix: vmw_vmci driver get_user_pages_fast error handling

[Mathieu Desnoyers]

Comparing a signed return value against an unsigned num_pages
field performs the comparison as "unsigned", and therefore mistakenly
considers get_user_pages_fast() errors as success.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Alexander Viro <viro@zeniv.linux.org.uk>
CC: George Zhang <georgezhang@vmware.com>
CC: Andy king <acking@vmware.com>
CC: Dmitry Torokhov <dtor@vmware.com>
CC: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
CC: linux-fsdevel@vger.kernel.org
---
 drivers/misc/vmw_vmci/vmci_queue_pair.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/misc/vmw_vmci/vmci_queue_pair.c b/drivers/misc/vmw_vmci/vmci_queue_pair.c
index 8af5c2672f71..f71db848ce81 100644
--- a/drivers/misc/vmw_vmci/vmci_queue_pair.c
+++ b/drivers/misc/vmw_vmci/vmci_queue_pair.c
@@ -755,7 +755,7 @@ static int qp_host_get_user_memory(u64 produce_uva,
 	retval = get_user_pages_fast((uintptr_t) produce_uva,
 				     produce_q->kernel_if->num_pages, 1,
 				     produce_q->kernel_if->u.h.header_page);
-	if (retval < produce_q->kernel_if->num_pages) {
+	if (retval < 0 || retval < produce_q->kernel_if->num_pages) {
 		pr_debug("get_user_pages_fast(produce) failed (retval=%d)",
 			retval);
 		qp_release_pages(produce_q->kernel_if->u.h.header_page,
@@ -767,7 +767,7 @@ static int qp_host_get_user_memory(u64 produce_uva,
 	retval = get_user_pages_fast((uintptr_t) consume_uva,
 				     consume_q->kernel_if->num_pages, 1,
 				     consume_q->kernel_if->u.h.header_page);
-	if (retval < consume_q->kernel_if->num_pages) {
+	if (retval < 0 || retval < consume_q->kernel_if->num_pages) {
 		pr_debug("get_user_pages_fast(consume) failed (retval=%d)",
 			retval);
 		qp_release_pages(consume_q->kernel_if->u.h.header_page,
-- 
2.11.0

Re: [PATCH 1/1] Fix: vmw_vmci driver get_user_pages_fast error handling

[Al Viro]

On Tue, Oct 31, 2017 at 07:00:38PM -0400, Mathieu Desnoyers wrote:
> Comparing a signed return value against an unsigned num_pages
> field performs the comparison as "unsigned", and therefore mistakenly
> considers get_user_pages_fast() errors as success.

It's worse than that - if you look into the code in question you'll
see
                pr_debug("get_user_pages_fast(produce) failed (retval=%d)",
                        retval);
                qp_release_pages(produce_q->kernel_if->u.h.header_page,
                                 retval, false);
                err = VMCI_ERROR_NO_MEM;
                goto out;
with
static void qp_release_pages(struct page **pages,
                             u64 num_pages, bool dirty)
{
        int i;

        for (i = 0; i < num_pages; i++) {
                if (dirty)
                        set_page_dirty(pages[i]);

                put_page(pages[i]);
                pages[i] = NULL;
        }
}

Now, guess what'll happen if you get there with retval being negative?
AFAICS, the right fix is something along the lines of
        if (retval != produce_q->kernel_if->num_pages) {
                pr_debug("get_user_pages_fast(produce) failed (retval=%d)",
                        retval);
		if (retval > 0)
			qp_release_pages(produce_q->kernel_if->u.h.header_page,
					 retval, false);
                err = VMCI_ERROR_NO_MEM;
                goto out;
        }
and similar for the second caller.  Objections?

Re: [PATCH 1/1] Fix: vmw_vmci driver get_user_pages_fast error handling

[Mathieu Desnoyers]

----- On Nov 2, 2017, at 2:46 PM, Alexander Viro viro@ZenIV.linux.org.uk wrote:

> On Tue, Oct 31, 2017 at 07:00:38PM -0400, Mathieu Desnoyers wrote:
>> Comparing a signed return value against an unsigned num_pages
>> field performs the comparison as "unsigned", and therefore mistakenly
>> considers get_user_pages_fast() errors as success.
> 
> It's worse than that - if you look into the code in question you'll
> see
>                pr_debug("get_user_pages_fast(produce) failed (retval=%d)",
>                        retval);
>                qp_release_pages(produce_q->kernel_if->u.h.header_page,
>                                 retval, false);
>                err = VMCI_ERROR_NO_MEM;
>                goto out;
> with
> static void qp_release_pages(struct page **pages,
>                             u64 num_pages, bool dirty)
> {
>        int i;
> 
>        for (i = 0; i < num_pages; i++) {
>                if (dirty)
>                        set_page_dirty(pages[i]);
> 
>                put_page(pages[i]);
>                pages[i] = NULL;
>        }
> }
> 
> Now, guess what'll happen if you get there with retval being negative?

Well this ought to be a pretty long loop...

> AFAICS, the right fix is something along the lines of
>        if (retval != produce_q->kernel_if->num_pages) {
>                pr_debug("get_user_pages_fast(produce) failed (retval=%d)",
>                        retval);
>		if (retval > 0)
>			qp_release_pages(produce_q->kernel_if->u.h.header_page,
>					 retval, false);
>                err = VMCI_ERROR_NO_MEM;
>                goto out;
>        }
> and similar for the second caller.  Objections?

No objection from me,

Thanks!

Mathieu


-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com