2016-12-03 10:51:19

by Pan Bian

Subject: [PATCH 1/1] xen: xenbus: set error code on failure

In function xenstored_local_init(), the value of return variable err
should be negative on errors. But the value of err keeps 0 even if the
call to get_zeroed_page() returns a NULL pointer. This patch assigns
"-ENOMEM" to err on the error branch.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=188721

Signed-off-by: Pan Bian <[email protected]>
---
drivers/xen/xenbus/xenbus_probe.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/xen/xenbus/xenbus_probe.c b/drivers/xen/xenbus/xenbus_probe.c
index 33a31cf..f87d047 100644
--- a/drivers/xen/xenbus/xenbus_probe.c
+++ b/drivers/xen/xenbus/xenbus_probe.c
@@ -708,8 +708,10 @@ static int __init xenstored_local_init(void)

/* Allocate Xenstore page */
page = get_zeroed_page(GFP_KERNEL);
- if (!page)
+ if (!page) {
+ err = -ENOMEM;
goto out_err;
+ }

xen_store_gfn = xen_start_info->store_mfn = virt_to_gfn((void *)page);

--
1.9.1

2016-12-05 06:30:54

by Jürgen Groß

Subject: Re: [PATCH 1/1] xen: xenbus: set error code on failure

On 03/12/16 11:49, Pan Bian wrote:
> In function xenstored_local_init(), the value of return variable err
> should be negative on errors. But the value of err keeps 0 even if the
> call to get_zeroed_page() returns a NULL pointer. This patch assigns
> "-ENOMEM" to err on the error branch.
>
> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=188721
>
> Signed-off-by: Pan Bian <[email protected]>
> ---
> drivers/xen/xenbus/xenbus_probe.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/xen/xenbus/xenbus_probe.c b/drivers/xen/xenbus/xenbus_probe.c
> index 33a31cf..f87d047 100644
> --- a/drivers/xen/xenbus/xenbus_probe.c
> +++ b/drivers/xen/xenbus/xenbus_probe.c
> @@ -708,8 +708,10 @@ static int __init xenstored_local_init(void)
>
> /* Allocate Xenstore page */
> page = get_zeroed_page(GFP_KERNEL);
> - if (!page)
> + if (!page) {
> + err = -ENOMEM;
> goto out_err;
> + }
>
> xen_store_gfn = xen_start_info->store_mfn = virt_to_gfn((void *)page);

Why don't you preset err to -ENOMEM instead? Initializing it to 0
is kind of pointless.

Ans while at it: preinitializing page isn't needed, too, and in the
error path testing page for being non-zero isn't neede either
(free_page() will do the right thing in case the parameter is 0).


Juergen

2016-12-05 07:54:03

by Pan Bian

Subject: Re: [PATCH 1/1] xen: xenbus: set error code on failure

From: PanBian <[email protected]>

On Mon, Dec 05, 2016 at 07:30:49AM +0100, Juergen Gross wrote:
> On 03/12/16 11:49, Pan Bian wrote:
> > In function xenstored_local_init(), the value of return variable err
> > should be negative on errors. But the value of err keeps 0 even if the
> > call to get_zeroed_page() returns a NULL pointer. This patch assigns
> > "-ENOMEM" to err on the error branch.
> >
> > Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=188721
> >
> > Signed-off-by: Pan Bian <[email protected]>
> > ---
> > drivers/xen/xenbus/xenbus_probe.c | 4 +++-
> > 1 file changed, 3 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/xen/xenbus/xenbus_probe.c b/drivers/xen/xenbus/xenbus_probe.c
> > index 33a31cf..f87d047 100644
> > --- a/drivers/xen/xenbus/xenbus_probe.c
> > +++ b/drivers/xen/xenbus/xenbus_probe.c
> > @@ -708,8 +708,10 @@ static int __init xenstored_local_init(void)
> >
> > /* Allocate Xenstore page */
> > page = get_zeroed_page(GFP_KERNEL);
> > - if (!page)
> > + if (!page) {
> > + err = -ENOMEM;
> > goto out_err;
> > + }
> >
> > xen_store_gfn = xen_start_info->store_mfn = virt_to_gfn((void *)page);
>
> Why don't you preset err to -ENOMEM instead? Initializing it to 0
> is kind of pointless.

I think presetting and setting on demand are both effective to fix
this bug. Nevertheless, I will resubmit a second version if you
insist.

>
> Ans while at it: preinitializing page isn't needed, too, and in the
> error path testing page for being non-zero isn't neede either
> (free_page() will do the right thing in case the parameter is 0).
>
>
> Juergen
>

Thanks!

Best regards,
Pan