| Message ID | 1310538499.4968.94.camel@pasglop (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, 2011-07-13 at 10:48 -0400, Alex Deucher wrote: > On Wed, Jul 13, 2011 at 10:43 AM, Alex Deucher <alexdeucher@gmail.com> wrote: > > On Wed, Jul 13, 2011 at 2:28 AM, Benjamin Herrenschmidt > > <benh@kernel.crashing.org> wrote: > >> We should have a read memory barrier between reading the WPTR from > >> memory and reading ring entries based on that value (ie, we need to > >> ensure both loads are done in order by the CPU). > >> > >> It could be argued that the MMIO reads in r600_ack_irq() might be > >> enough to get that barrier but I prefer keeping an explicit one just > >> in case. > >> > >> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> > > > > Reviewed-by: Alex Deucher <alexander.deucher@amd.com> > > evergreen.c will need a similar fix. Ok. I can do that. Cheers, Ben. > Alex > > > > >> --- > >> > >> (resent adding dri-devel to the CC list to hit patchwork) > >> > >> drivers/gpu/drm/radeon/r600.c | 3 +++ > >> 1 files changed, 3 insertions(+), 0 deletions(-) > >> > >> diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c > >> index 3c86b15..7e5c801 100644 > >> --- a/drivers/gpu/drm/radeon/r600.c > >> +++ b/drivers/gpu/drm/radeon/r600.c > >> @@ -3312,6 +3312,9 @@ int r600_irq_process(struct radeon_device *rdev) > >> } > >> > >> restart_ih: > >> + /* Order reading of wptr vs. reading of IH ring data */ > >> + wmb(); > >> + > >> /* display interrupts */ > >> r600_irq_ack(rdev); > >> > >> > >> > >> > >> > >
On Wed, Jul 13, 2011 at 5:42 PM, Benjamin Herrenschmidt <benh@kernel.crashing.org> wrote: > On Wed, 2011-07-13 at 10:48 -0400, Alex Deucher wrote: >> On Wed, Jul 13, 2011 at 10:43 AM, Alex Deucher <alexdeucher@gmail.com> wrote: >> > On Wed, Jul 13, 2011 at 2:28 AM, Benjamin Herrenschmidt >> > <benh@kernel.crashing.org> wrote: >> >> We should have a read memory barrier between reading the WPTR from >> >> memory and reading ring entries based on that value (ie, we need to >> >> ensure both loads are done in order by the CPU). >> >> >> >> It could be argued that the MMIO reads in r600_ack_irq() might be >> >> enough to get that barrier but I prefer keeping an explicit one just >> >> in case. >> >> >> >> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> >> > >> > Reviewed-by: Alex Deucher <alexander.deucher@amd.com> >> >> evergreen.c will need a similar fix. > > Ok. I can do that. Thanks! > > Cheers, > Ben. > >> Alex >> >> > >> >> --- >> >> >> >> (resent adding dri-devel to the CC list to hit patchwork) >> >> >> >> drivers/gpu/drm/radeon/r600.c | 3 +++ >> >> 1 files changed, 3 insertions(+), 0 deletions(-) >> >> >> >> diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c >> >> index 3c86b15..7e5c801 100644 >> >> --- a/drivers/gpu/drm/radeon/r600.c >> >> +++ b/drivers/gpu/drm/radeon/r600.c >> >> @@ -3312,6 +3312,9 @@ int r600_irq_process(struct radeon_device *rdev) >> >> } >> >> >> >> restart_ih: >> >> + /* Order reading of wptr vs. reading of IH ring data */ >> >> + wmb(); >> >> + >> >> /* display interrupts */ >> >> r600_irq_ack(rdev); >> >> >> >> >> >> >> >> >> >> >> > > > >
On Wed, Jul 13, 2011 at 6:28 AM, Benjamin Herrenschmidt <benh@kernel.crashing.org> wrote: > We should have a read memory barrier between reading the WPTR from > memory and reading ring entries based on that value (ie, we need to > ensure both loads are done in order by the CPU). > > It could be argued that the MMIO reads in r600_ack_irq() might be > enough to get that barrier but I prefer keeping an explicit one just > in case. > > Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> > --- > > (resent adding dri-devel to the CC list to hit patchwork) > > drivers/gpu/drm/radeon/r600.c | 3 +++ > 1 files changed, 3 insertions(+), 0 deletions(-) > > diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c > index 3c86b15..7e5c801 100644 > --- a/drivers/gpu/drm/radeon/r600.c > +++ b/drivers/gpu/drm/radeon/r600.c > @@ -3312,6 +3312,9 @@ int r600_irq_process(struct radeon_device *rdev) > } > > restart_ih: > + /* Order reading of wptr vs. reading of IH ring data */ > + wmb(); > + > /* display interrupts */ > r600_irq_ack(rdev); The subject line says rmb(), but this says wmb(). Just want to verify what you have is correct. Matt
On Fri, 2011-07-15 at 04:19 +0000, Matt Turner wrote: > On Wed, Jul 13, 2011 at 6:28 AM, Benjamin Herrenschmidt > <benh@kernel.crashing.org> wrote: > > We should have a read memory barrier between reading the WPTR from > > memory and reading ring entries based on that value (ie, we need to > > ensure both loads are done in order by the CPU). > > > > It could be argued that the MMIO reads in r600_ack_irq() might be > > enough to get that barrier but I prefer keeping an explicit one just > > in case. > > > > Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> > > --- > > > > (resent adding dri-devel to the CC list to hit patchwork) > > > > drivers/gpu/drm/radeon/r600.c | 3 +++ > > 1 files changed, 3 insertions(+), 0 deletions(-) > > > > diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c > > index 3c86b15..7e5c801 100644 > > --- a/drivers/gpu/drm/radeon/r600.c > > +++ b/drivers/gpu/drm/radeon/r600.c > > @@ -3312,6 +3312,9 @@ int r600_irq_process(struct radeon_device *rdev) > > } > > > > restart_ih: > > + /* Order reading of wptr vs. reading of IH ring data */ > > + wmb(); > > + > > /* display interrupts */ > > r600_irq_ack(rdev); > > The subject line says rmb(), but this says wmb(). Just want to verify > what you have is correct. Nice spotting, it's a typo and should have been rmb(). I'll fix it and respin. Cheers, Ben
diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 3c86b15..7e5c801 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -3312,6 +3312,9 @@ int r600_irq_process(struct radeon_device *rdev) } restart_ih: + /* Order reading of wptr vs. reading of IH ring data */ + wmb(); + /* display interrupts */ r600_irq_ack(rdev);
We should have a read memory barrier between reading the WPTR from memory and reading ring entries based on that value (ie, we need to ensure both loads are done in order by the CPU). It could be argued that the MMIO reads in r600_ack_irq() might be enough to get that barrier but I prefer keeping an explicit one just in case. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- (resent adding dri-devel to the CC list to hit patchwork) drivers/gpu/drm/radeon/r600.c | 3 +++ 1 files changed, 3 insertions(+), 0 deletions(-)