| Message ID | 25470133.K8n9sLBzRS@diego (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi Heiko, Sandy, On Fri, May 25, 2018 at 7:07 AM Heiko Stübner <heiko@sntech.de> wrote: > From: Sandy Huang <hjc@rock-chips.com> > The vop irq is shared between vop and iommu and irq probing in the > iommu driver moved to the probe function recently. This can in some > cases lead to a stall if the irq is triggered while the vop driver > still has it disabled. > But there is no real need to disable the irq, as the vop can simply > also track its enabled state and ignore irqs in that case. > So remove the enable/disable handling and add appropriate condition > to the irq handler. > Signed-off-by: Sandy Huang <hjc@rock-chips.com> > [added an actual commit message] > Signed-off-by: Heiko Stuebner <heiko@sntech.de> > --- > Hi Ezequiel, > this patch came from a discussion I had with Rockchip people over the > iommu changes and resulting issues back in april, but somehow was > forgotten and not posted to the lists. Correcting that now. > So removing the enable/disable voodoo on the shared interrupt is > the preferred way. > drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 13 ++++++------- > 1 file changed, 7 insertions(+), 7 deletions(-) > diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c > index 510cdf0..61493d4 100644 > --- a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c > +++ b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c > @@ -549,8 +549,6 @@ static int vop_enable(struct drm_crtc *crtc) > spin_unlock(&vop->reg_lock); > - enable_irq(vop->irq); > - While this one should be okay (+/- my comment for vop_isr()), because the hardware is already powered on and clocked at this point... > drm_crtc_vblank_on(crtc); > return 0; > @@ -596,8 +594,6 @@ static void vop_crtc_atomic_disable(struct drm_crtc *crtc, > vop_dsp_hold_valid_irq_disable(vop); > - disable_irq(vop->irq); > - > vop->is_enabled = false; ...this one is more tricky. There might be an interrupt handler still running at this point. disable_irq() waits for any running handler to complete before disabling, so we might want to call synchronize_irq() after setting is_enabled to false. > /* > @@ -1168,6 +1164,13 @@ static irqreturn_t vop_isr(int irq, void *data) > int ret = IRQ_NONE; > /* > + * since the irq is shared with iommu, iommu irq is enabled before vop > + * enable, so before vop enable we do nothing. > + */ > + if (!vop->is_enabled) > + return IRQ_NONE; This doesn't seem to be properly synchronized. We don't even call it under a spinlock, so no barriers are issued. Perhaps we should make this atomic_t? Best regards, Tomasz
[Thanks Robin for pointing me to that patch.] Hi Heiko, On 24/05/18 23:06, Heiko Stübner wrote: > From: Sandy Huang <hjc@rock-chips.com> > > The vop irq is shared between vop and iommu and irq probing in the > iommu driver moved to the probe function recently. This can in some > cases lead to a stall if the irq is triggered while the vop driver > still has it disabled. > > But there is no real need to disable the irq, as the vop can simply > also track its enabled state and ignore irqs in that case. > > So remove the enable/disable handling and add appropriate condition > to the irq handler. > > Signed-off-by: Sandy Huang <hjc@rock-chips.com> > [added an actual commit message] > Signed-off-by: Heiko Stuebner <heiko@sntech.de> > --- > Hi Ezequiel, > > this patch came from a discussion I had with Rockchip people over the > iommu changes and resulting issues back in april, but somehow was > forgotten and not posted to the lists. Correcting that now. > > So removing the enable/disable voodoo on the shared interrupt is > the preferred way. > > drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 13 ++++++------- > 1 file changed, 7 insertions(+), 7 deletions(-) > > diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c > index 510cdf0..61493d4 100644 > --- a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c > +++ b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c > @@ -549,8 +549,6 @@ static int vop_enable(struct drm_crtc *crtc) > > spin_unlock(&vop->reg_lock); > > - enable_irq(vop->irq); > - > drm_crtc_vblank_on(crtc); > > return 0; > @@ -596,8 +594,6 @@ static void vop_crtc_atomic_disable(struct drm_crtc *crtc, > > vop_dsp_hold_valid_irq_disable(vop); > > - disable_irq(vop->irq); > - > vop->is_enabled = false; > > /* > @@ -1168,6 +1164,13 @@ static irqreturn_t vop_isr(int irq, void *data) > int ret = IRQ_NONE; > > /* > + * since the irq is shared with iommu, iommu irq is enabled before vop > + * enable, so before vop enable we do nothing. > + */ > + if (!vop->is_enabled) > + return IRQ_NONE; > + What guarantee do we have that the IOMMU will actually service that interrupt? Can't we get into a situation where the interrupt gets ignored by both drivers and subsequently disabled by the core irq code as being spurious? From what I understood of the way things are plugged together on the rockchip platforms, each individual VOP is behind a single IOMMU, hence there there is hardly any point in handling IOMMU interrupts if the VOP is off. But I'm missing the actual reason behind this patch. Could you enlighten me? Thanks, M.
Hi Marc, Am Freitag, 25. Mai 2018, 13:07:02 CEST schrieb Marc Zyngier: > [Thanks Robin for pointing me to that patch.] > > Hi Heiko, > > On 24/05/18 23:06, Heiko Stübner wrote: > > From: Sandy Huang <hjc@rock-chips.com> > > > > The vop irq is shared between vop and iommu and irq probing in the > > iommu driver moved to the probe function recently. This can in some > > cases lead to a stall if the irq is triggered while the vop driver > > still has it disabled. > > > > But there is no real need to disable the irq, as the vop can simply > > also track its enabled state and ignore irqs in that case. > > > > So remove the enable/disable handling and add appropriate condition > > to the irq handler. > > > > Signed-off-by: Sandy Huang <hjc@rock-chips.com> > > [added an actual commit message] > > Signed-off-by: Heiko Stuebner <heiko@sntech.de> > > --- > > Hi Ezequiel, > > > > this patch came from a discussion I had with Rockchip people over the > > iommu changes and resulting issues back in april, but somehow was > > forgotten and not posted to the lists. Correcting that now. > > > > So removing the enable/disable voodoo on the shared interrupt is > > the preferred way. > > > > drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 13 ++++++------- > > 1 file changed, 7 insertions(+), 7 deletions(-) > > > > diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c > > index 510cdf0..61493d4 100644 > > --- a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c > > +++ b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c > > @@ -549,8 +549,6 @@ static int vop_enable(struct drm_crtc *crtc) > > > > spin_unlock(&vop->reg_lock); > > > > - enable_irq(vop->irq); > > - > > drm_crtc_vblank_on(crtc); > > > > return 0; > > @@ -596,8 +594,6 @@ static void vop_crtc_atomic_disable(struct drm_crtc *crtc, > > > > vop_dsp_hold_valid_irq_disable(vop); > > > > - disable_irq(vop->irq); > > - > > vop->is_enabled = false; > > > > /* > > @@ -1168,6 +1164,13 @@ static irqreturn_t vop_isr(int irq, void *data) > > int ret = IRQ_NONE; > > > > /* > > + * since the irq is shared with iommu, iommu irq is enabled before vop > > + * enable, so before vop enable we do nothing. > > + */ > > + if (!vop->is_enabled) > > + return IRQ_NONE; > > + > > What guarantee do we have that the IOMMU will actually service that > interrupt? Can't we get into a situation where the interrupt gets > ignored by both drivers and subsequently disabled by the core irq code > as being spurious? > > From what I understood of the way things are plugged together on the > rockchip platforms, each individual VOP is behind a single IOMMU, hence > there there is hardly any point in handling IOMMU interrupts if the VOP > is off. Yeah, which is probably the reason that patch fell through the cracks initially. I resurrected it from the internal discussion because of the issue described below. > But I'm missing the actual reason behind this patch. Could you enlighten me? Recently the iommu driver moved the irq request from the iommu enablement to its probe function. With that change Ezequiel got various stalls, see https://lkml.org/lkml/2018/5/24/1347 Heiko
diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c index 510cdf0..61493d4 100644 --- a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c +++ b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c @@ -549,8 +549,6 @@ static int vop_enable(struct drm_crtc *crtc) spin_unlock(&vop->reg_lock); - enable_irq(vop->irq); - drm_crtc_vblank_on(crtc); return 0; @@ -596,8 +594,6 @@ static void vop_crtc_atomic_disable(struct drm_crtc *crtc, vop_dsp_hold_valid_irq_disable(vop); - disable_irq(vop->irq); - vop->is_enabled = false; /* @@ -1168,6 +1164,13 @@ static irqreturn_t vop_isr(int irq, void *data) int ret = IRQ_NONE; /* + * since the irq is shared with iommu, iommu irq is enabled before vop + * enable, so before vop enable we do nothing. + */ + if (!vop->is_enabled) + return IRQ_NONE; + + /* * interrupt register has interrupt status, enable and clear bits, we * must hold irq_lock to avoid a race with enable/disable_vblank(). */ @@ -1586,9 +1588,6 @@ static int vop_bind(struct device *dev, struct device *master, void *data) if (ret) goto err_disable_pm_runtime; - /* IRQ is initially disabled; it gets enabled in power_on */ - disable_irq(vop->irq); - return 0; err_disable_pm_runtime: