ppc/iommu: prevent false TCE leak message

When a device DMA window includes the address 0, it's reserved in the
TCE bitmap to avoid returning that address to drivers.

When the device is removed, the bitmap is checked for any mappings not
removed by the driver, indicating a possible DMA mapping leak. Since the
reserved address is not cleared, a message is printed, warning of such a
leak.

Check for the reservation, and clear it before checking for any other
standing mappings.

Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@linux.vnet.ibm.com>
---
 arch/powerpc/kernel/iommu.c |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)

On Thu, Dec 27, 2012 at 02:28:06PM -0200, Thadeu Lima de Souza Cascardo wrote:
>When a device DMA window includes the address 0, it's reserved in the
>TCE bitmap to avoid returning that address to drivers.
>
>When the device is removed, the bitmap is checked for any mappings not
>removed by the driver, indicating a possible DMA mapping leak. Since the
>reserved address is not cleared, a message is printed, warning of such a
>leak.
>
>Check for the reservation, and clear it before checking for any other
>standing mappings.
>
>Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@linux.vnet.ibm.com>
>---
> arch/powerpc/kernel/iommu.c |    5 +++++
> 1 files changed, 5 insertions(+), 0 deletions(-)
>
>diff --git a/arch/powerpc/kernel/iommu.c b/arch/powerpc/kernel/iommu.c
>index 8226c6c..226e9e5 100644
>--- a/arch/powerpc/kernel/iommu.c
>+++ b/arch/powerpc/kernel/iommu.c
>@@ -717,6 +717,11 @@ void iommu_free_table(struct iommu_table *tbl, const char *node_name)
> 		return;
> 	}
>
>+	/* In case we have reserved the first bit, we should not emit
>+	 * the warning below. */

At least, the comment would look like:

	/*
	 * xxxxxxx
	 */

>+	if (tbl->it_offset == 0)
>+		clear_bit(0, tbl->it_map);
>+
> 	/* verify that table contains no entries */
> 	/* it_size is in entries, and we're examining 64 at a time */

The comment would be merged as well? :-)

> 	for (i = 0; i < (tbl->it_size/64); i++) {

Thanks,
Gavin

On Fri, Dec 28, 2012 at 01:21:35PM +0800, Gavin Shan wrote:
> On Thu, Dec 27, 2012 at 02:28:06PM -0200, Thadeu Lima de Souza Cascardo wrote:
> >When a device DMA window includes the address 0, it's reserved in the
> >TCE bitmap to avoid returning that address to drivers.
> >
> >When the device is removed, the bitmap is checked for any mappings not
> >removed by the driver, indicating a possible DMA mapping leak. Since the
> >reserved address is not cleared, a message is printed, warning of such a
> >leak.
> >
> >Check for the reservation, and clear it before checking for any other
> >standing mappings.
> >
> >Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@linux.vnet.ibm.com>
> >---
> > arch/powerpc/kernel/iommu.c |    5 +++++
> > 1 files changed, 5 insertions(+), 0 deletions(-)
> >
> >diff --git a/arch/powerpc/kernel/iommu.c b/arch/powerpc/kernel/iommu.c
> >index 8226c6c..226e9e5 100644
> >--- a/arch/powerpc/kernel/iommu.c
> >+++ b/arch/powerpc/kernel/iommu.c
> >@@ -717,6 +717,11 @@ void iommu_free_table(struct iommu_table *tbl, const char *node_name)
> > 		return;
> > 	}
> >
> >+	/* In case we have reserved the first bit, we should not emit
> >+	 * the warning below. */
> 
> At least, the comment would look like:
> 
> 	/*
> 	 * xxxxxxx
> 	 */
> 

Sure. New code should always follow coding style.  :-)

How do you suggest merging with the comment below? I think it's closer
to the code it comments about, so I'd rather keep it where it is.

> >+	if (tbl->it_offset == 0)
> >+		clear_bit(0, tbl->it_map);
> >+
> > 	/* verify that table contains no entries */
> > 	/* it_size is in entries, and we're examining 64 at a time */
> 
> The comment would be merged as well? :-)
> 

I also considered replacing this code by this:

        /* verify that table contains no entries */
-       /* it_size is in entries, and we're examining 64 at a time */
-       for (i = 0; i < (tbl->it_size/64); i++) {
-               if (tbl->it_map[i] != 0) {
+       if (find_first_bit(tbl->it_map, tbl->it_size) < tbl->it_size)
                        printk(KERN_WARNING "%s: Unexpected TCEs for
%s\n",
                                __func__, node_name);
-                       break;
-               }
-       }
 
I'll resend the "unreserving" patch with the fixed comment styling, but
without merging comments. And I will send this other patch for comments.

Regards.
Thadeu Cascardo.

 
> > 	for (i = 0; i < (tbl->it_size/64); i++) {
> 
> Thanks,
> Gavin