With the submission of iommu driver for RK3568 a subtle bug was
introduced: PAGE_DESC_HI_MASK1 and PAGE_DESC_HI_MASK2 have to be
the other way arround - that leads to random errors, especially when
addresses beyond 32 bit are used.
Fix it.
Fixes: c55356c534aa ("iommu: rockchip: Add support for iommu v2")
Signed-off-by: Alex Bee <[email protected]>
---
I've found this bug, when testing the recent VOP2 submission, which is
the first perpherial which uses iommu for RK356x. I could use it on my
quartz64 8GB board only, when limiting the available memory to less 4GB.
This patch fixes it.
drivers/iommu/rockchip-iommu.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/iommu/rockchip-iommu.c b/drivers/iommu/rockchip-iommu.c
index 5cb260820eda..7f23ad61c094 100644
--- a/drivers/iommu/rockchip-iommu.c
+++ b/drivers/iommu/rockchip-iommu.c
@@ -200,8 +200,8 @@ static inline phys_addr_t rk_dte_pt_address(u32 dte)
#define DTE_HI_MASK2 GENMASK(7, 4)
#define DTE_HI_SHIFT1 24 /* shift bit 8 to bit 32 */
#define DTE_HI_SHIFT2 32 /* shift bit 4 to bit 36 */
-#define PAGE_DESC_HI_MASK1 GENMASK_ULL(39, 36)
-#define PAGE_DESC_HI_MASK2 GENMASK_ULL(35, 32)
+#define PAGE_DESC_HI_MASK1 GENMASK_ULL(35, 32)
+#define PAGE_DESC_HI_MASK2 GENMASK_ULL(39, 36)
static inline phys_addr_t rk_dte_pt_address_v2(u32 dte)
{
--
2.30.2
Le 24/11/2021 à 03:13, Alex Bee a écrit :
> With the submission of iommu driver for RK3568 a subtle bug was
> introduced: PAGE_DESC_HI_MASK1 and PAGE_DESC_HI_MASK2 have to be
> the other way arround - that leads to random errors, especially when
> addresses beyond 32 bit are used.
>
> Fix it.
>
> Fixes: c55356c534aa ("iommu: rockchip: Add support for iommu v2")
> Signed-off-by: Alex Bee <[email protected]>
Thanks for the fix.
Reviewed-by: Benjamin Gaignard <[email protected]>
> ---
>
> I've found this bug, when testing the recent VOP2 submission, which is
> the first perpherial which uses iommu for RK356x. I could use it on my
> quartz64 8GB board only, when limiting the available memory to less 4GB.
>
> This patch fixes it.
>
> drivers/iommu/rockchip-iommu.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/iommu/rockchip-iommu.c b/drivers/iommu/rockchip-iommu.c
> index 5cb260820eda..7f23ad61c094 100644
> --- a/drivers/iommu/rockchip-iommu.c
> +++ b/drivers/iommu/rockchip-iommu.c
> @@ -200,8 +200,8 @@ static inline phys_addr_t rk_dte_pt_address(u32 dte)
> #define DTE_HI_MASK2 GENMASK(7, 4)
> #define DTE_HI_SHIFT1 24 /* shift bit 8 to bit 32 */
> #define DTE_HI_SHIFT2 32 /* shift bit 4 to bit 36 */
> -#define PAGE_DESC_HI_MASK1 GENMASK_ULL(39, 36)
> -#define PAGE_DESC_HI_MASK2 GENMASK_ULL(35, 32)
> +#define PAGE_DESC_HI_MASK1 GENMASK_ULL(35, 32)
> +#define PAGE_DESC_HI_MASK2 GENMASK_ULL(39, 36)
>
> static inline phys_addr_t rk_dte_pt_address_v2(u32 dte)
> {
Am Mittwoch, 24. November 2021, 03:13:25 CET schrieb Alex Bee:
> With the submission of iommu driver for RK3568 a subtle bug was
> introduced: PAGE_DESC_HI_MASK1 and PAGE_DESC_HI_MASK2 have to be
> the other way arround - that leads to random errors, especially when
> addresses beyond 32 bit are used.
>
> Fix it.
>
> Fixes: c55356c534aa ("iommu: rockchip: Add support for iommu v2")
> Signed-off-by: Alex Bee <[email protected]>
> ---
>
> I've found this bug, when testing the recent VOP2 submission, which is
> the first perpherial which uses iommu for RK356x. I could use it on my
> quartz64 8GB board only, when limiting the available memory to less 4GB.
>
> This patch fixes it.
>
> drivers/iommu/rockchip-iommu.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/iommu/rockchip-iommu.c b/drivers/iommu/rockchip-iommu.c
> index 5cb260820eda..7f23ad61c094 100644
> --- a/drivers/iommu/rockchip-iommu.c
> +++ b/drivers/iommu/rockchip-iommu.c
> @@ -200,8 +200,8 @@ static inline phys_addr_t rk_dte_pt_address(u32 dte)
> #define DTE_HI_MASK2 GENMASK(7, 4)
> #define DTE_HI_SHIFT1 24 /* shift bit 8 to bit 32 */
> #define DTE_HI_SHIFT2 32 /* shift bit 4 to bit 36 */
> -#define PAGE_DESC_HI_MASK1 GENMASK_ULL(39, 36)
> -#define PAGE_DESC_HI_MASK2 GENMASK_ULL(35, 32)
> +#define PAGE_DESC_HI_MASK1 GENMASK_ULL(35, 32)
> +#define PAGE_DESC_HI_MASK2 GENMASK_ULL(39, 36)
With this change the PAGE_DESC_HI_MASK* now match the other
constants above them in what bits they access, so
Reviewed-by: Heiko Stuebner <[email protected]>
On Wed, Nov 24, 2021 at 5:36 AM Heiko Stübner <[email protected]> wrote:
>
> Am Mittwoch, 24. November 2021, 03:13:25 CET schrieb Alex Bee:
> > With the submission of iommu driver for RK3568 a subtle bug was
> > introduced: PAGE_DESC_HI_MASK1 and PAGE_DESC_HI_MASK2 have to be
> > the other way arround - that leads to random errors, especially when
> > addresses beyond 32 bit are used.
> >
> > Fix it.
> >
> > Fixes: c55356c534aa ("iommu: rockchip: Add support for iommu v2")
> > Signed-off-by: Alex Bee <[email protected]>
> > ---
> >
> > I've found this bug, when testing the recent VOP2 submission, which is
> > the first perpherial which uses iommu for RK356x. I could use it on my
> > quartz64 8GB board only, when limiting the available memory to less 4GB.
> >
> > This patch fixes it.
> >
> > drivers/iommu/rockchip-iommu.c | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/iommu/rockchip-iommu.c b/drivers/iommu/rockchip-iommu.c
> > index 5cb260820eda..7f23ad61c094 100644
> > --- a/drivers/iommu/rockchip-iommu.c
> > +++ b/drivers/iommu/rockchip-iommu.c
> > @@ -200,8 +200,8 @@ static inline phys_addr_t rk_dte_pt_address(u32 dte)
> > #define DTE_HI_MASK2 GENMASK(7, 4)
> > #define DTE_HI_SHIFT1 24 /* shift bit 8 to bit 32 */
> > #define DTE_HI_SHIFT2 32 /* shift bit 4 to bit 36 */
> > -#define PAGE_DESC_HI_MASK1 GENMASK_ULL(39, 36)
> > -#define PAGE_DESC_HI_MASK2 GENMASK_ULL(35, 32)
> > +#define PAGE_DESC_HI_MASK1 GENMASK_ULL(35, 32)
> > +#define PAGE_DESC_HI_MASK2 GENMASK_ULL(39, 36)
>
> With this change the PAGE_DESC_HI_MASK* now match the other
> constants above them in what bits they access, so
>
> Reviewed-by: Heiko Stuebner <[email protected]>
>
>
Tested on the quartz64-a board with the experimental vop2 driver,
confirmed it fixed the issue.
Tested-by: Peter Geis <[email protected]>
>
>
>
> _______________________________________________
> Linux-rockchip mailing list
> [email protected]
> http://lists.infradead.org/mailman/listinfo/linux-rockchip
On Wed, Nov 24, 2021 at 03:13:25AM +0100, Alex Bee wrote:
> With the submission of iommu driver for RK3568 a subtle bug was
> introduced: PAGE_DESC_HI_MASK1 and PAGE_DESC_HI_MASK2 have to be
> the other way arround - that leads to random errors, especially when
> addresses beyond 32 bit are used.
>
> Fix it.
>
> Fixes: c55356c534aa ("iommu: rockchip: Add support for iommu v2")
> Signed-off-by: Alex Bee <[email protected]>
Applied for v5.16, thanks.