| Message ID | 1442283858-27254-1-git-send-email-matthew.d.roper@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Tue, 15 Sep 2015, Matt Roper <matthew.d.roper@intel.com> wrote: > We allocate memory for LVDS modes while parsing the VBT at startup, but > never free this memory when the driver is unloaded, causing a small > leak. > > Signed-off-by: Matt Roper <matthew.d.roper@intel.com> Reviewed-by: Jani Nikula <jani.nikula@intel.com> We could probably use devm_*alloc functions for this kind of stuff more. > --- > drivers/gpu/drm/i915/i915_dma.c | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c > index 066a0ef..c069550 100644 > --- a/drivers/gpu/drm/i915/i915_dma.c > +++ b/drivers/gpu/drm/i915/i915_dma.c > @@ -1137,6 +1137,10 @@ int i915_driver_unload(struct drm_device *dev) > dev_priv->vbt.child_dev = NULL; > dev_priv->vbt.child_dev_num = 0; > } > + kfree(dev_priv->vbt.sdvo_lvds_vbt_mode); > + dev_priv->vbt.sdvo_lvds_vbt_mode = NULL; > + kfree(dev_priv->vbt.lfp_lvds_vbt_mode); > + dev_priv->vbt.lfp_lvds_vbt_mode = NULL; > > vga_switcheroo_unregister_client(dev->pdev); > vga_client_register(dev->pdev, NULL, NULL, NULL); > -- > 2.1.4 > > _______________________________________________ > Intel-gfx mailing list > Intel-gfx@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
On Tue, 15 Sep 2015, Matt Roper <matthew.d.roper@intel.com> wrote: > We allocate memory for LVDS modes while parsing the VBT at startup, but > never free this memory when the driver is unloaded, causing a small > leak. > > Signed-off-by: Matt Roper <matthew.d.roper@intel.com> Pushed to drm-intel-next-queued, thanks for the patch. BR, Jani. > --- > drivers/gpu/drm/i915/i915_dma.c | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c > index 066a0ef..c069550 100644 > --- a/drivers/gpu/drm/i915/i915_dma.c > +++ b/drivers/gpu/drm/i915/i915_dma.c > @@ -1137,6 +1137,10 @@ int i915_driver_unload(struct drm_device *dev) > dev_priv->vbt.child_dev = NULL; > dev_priv->vbt.child_dev_num = 0; > } > + kfree(dev_priv->vbt.sdvo_lvds_vbt_mode); > + dev_priv->vbt.sdvo_lvds_vbt_mode = NULL; > + kfree(dev_priv->vbt.lfp_lvds_vbt_mode); > + dev_priv->vbt.lfp_lvds_vbt_mode = NULL; > > vga_switcheroo_unregister_client(dev->pdev); > vga_client_register(dev->pdev, NULL, NULL, NULL); > -- > 2.1.4 > > _______________________________________________ > Intel-gfx mailing list > Intel-gfx@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
On Tue, Sep 15, 2015 at 10:50:45AM +0300, Jani Nikula wrote: > On Tue, 15 Sep 2015, Matt Roper <matthew.d.roper@intel.com> wrote: > > We allocate memory for LVDS modes while parsing the VBT at startup, but > > never free this memory when the driver is unloaded, causing a small > > leak. > > > > Signed-off-by: Matt Roper <matthew.d.roper@intel.com> > > Reviewed-by: Jani Nikula <jani.nikula@intel.com> > > We could probably use devm_*alloc functions for this kind of stuff more. The problem with devm_* is that the lifetime of the underlying hw device doesn't stricltly match the lifetime of the drm device. Not that anything in drm gets that right (pretty far from it), but that's why I haven't gone ballastic yet with rolling out devm_* all over our init code. -Daniel > > > > --- > > drivers/gpu/drm/i915/i915_dma.c | 4 ++++ > > 1 file changed, 4 insertions(+) > > > > diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c > > index 066a0ef..c069550 100644 > > --- a/drivers/gpu/drm/i915/i915_dma.c > > +++ b/drivers/gpu/drm/i915/i915_dma.c > > @@ -1137,6 +1137,10 @@ int i915_driver_unload(struct drm_device *dev) > > dev_priv->vbt.child_dev = NULL; > > dev_priv->vbt.child_dev_num = 0; > > } > > + kfree(dev_priv->vbt.sdvo_lvds_vbt_mode); > > + dev_priv->vbt.sdvo_lvds_vbt_mode = NULL; > > + kfree(dev_priv->vbt.lfp_lvds_vbt_mode); > > + dev_priv->vbt.lfp_lvds_vbt_mode = NULL; > > > > vga_switcheroo_unregister_client(dev->pdev); > > vga_client_register(dev->pdev, NULL, NULL, NULL); > > -- > > 2.1.4 > > > > _______________________________________________ > > Intel-gfx mailing list > > Intel-gfx@lists.freedesktop.org > > http://lists.freedesktop.org/mailman/listinfo/intel-gfx > > -- > Jani Nikula, Intel Open Source Technology Center > _______________________________________________ > Intel-gfx mailing list > Intel-gfx@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
On Wed, 23 Sep 2015, Daniel Vetter <daniel@ffwll.ch> wrote: > On Tue, Sep 15, 2015 at 10:50:45AM +0300, Jani Nikula wrote: >> On Tue, 15 Sep 2015, Matt Roper <matthew.d.roper@intel.com> wrote: >> > We allocate memory for LVDS modes while parsing the VBT at startup, but >> > never free this memory when the driver is unloaded, causing a small >> > leak. >> > >> > Signed-off-by: Matt Roper <matthew.d.roper@intel.com> >> >> Reviewed-by: Jani Nikula <jani.nikula@intel.com> >> >> We could probably use devm_*alloc functions for this kind of stuff more. > > The problem with devm_* is that the lifetime of the underlying hw > device doesn't stricltly match the lifetime of the drm device. Is it not enough that the lifetime of the driver is bounded by the lifetime of the drm device? What is the problem you're referring to, apart from handing the hardware devm allocated memory and unreferencing the drm device? > Not that anything in drm gets that right (pretty far from it), but > that's why I haven't gone ballastic yet with rolling out devm_* all > over our init code. I think you mean "ballistic", but, in this case, I think your typo "ballastic" is actually a more accurate description of the change. :) BR, Jani. > -Daniel > >> >> >> > --- >> > drivers/gpu/drm/i915/i915_dma.c | 4 ++++ >> > 1 file changed, 4 insertions(+) >> > >> > diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c >> > index 066a0ef..c069550 100644 >> > --- a/drivers/gpu/drm/i915/i915_dma.c >> > +++ b/drivers/gpu/drm/i915/i915_dma.c >> > @@ -1137,6 +1137,10 @@ int i915_driver_unload(struct drm_device *dev) >> > dev_priv->vbt.child_dev = NULL; >> > dev_priv->vbt.child_dev_num = 0; >> > } >> > + kfree(dev_priv->vbt.sdvo_lvds_vbt_mode); >> > + dev_priv->vbt.sdvo_lvds_vbt_mode = NULL; >> > + kfree(dev_priv->vbt.lfp_lvds_vbt_mode); >> > + dev_priv->vbt.lfp_lvds_vbt_mode = NULL; >> > >> > vga_switcheroo_unregister_client(dev->pdev); >> > vga_client_register(dev->pdev, NULL, NULL, NULL); >> > -- >> > 2.1.4 >> > >> > _______________________________________________ >> > Intel-gfx mailing list >> > Intel-gfx@lists.freedesktop.org >> > http://lists.freedesktop.org/mailman/listinfo/intel-gfx >> >> -- >> Jani Nikula, Intel Open Source Technology Center >> _______________________________________________ >> Intel-gfx mailing list >> Intel-gfx@lists.freedesktop.org >> http://lists.freedesktop.org/mailman/listinfo/intel-gfx > > -- > Daniel Vetter > Software Engineer, Intel Corporation > http://blog.ffwll.ch
On Wed, Sep 23, 2015 at 02:02:03PM +0300, Jani Nikula wrote: > On Wed, 23 Sep 2015, Daniel Vetter <daniel@ffwll.ch> wrote: > > On Tue, Sep 15, 2015 at 10:50:45AM +0300, Jani Nikula wrote: > >> On Tue, 15 Sep 2015, Matt Roper <matthew.d.roper@intel.com> wrote: > >> > We allocate memory for LVDS modes while parsing the VBT at startup, but > >> > never free this memory when the driver is unloaded, causing a small > >> > leak. > >> > > >> > Signed-off-by: Matt Roper <matthew.d.roper@intel.com> > >> > >> Reviewed-by: Jani Nikula <jani.nikula@intel.com> > >> > >> We could probably use devm_*alloc functions for this kind of stuff more. > > > > The problem with devm_* is that the lifetime of the underlying hw > > device doesn't stricltly match the lifetime of the drm device. > > Is it not enough that the lifetime of the driver is bounded by the > lifetime of the drm device? What is the problem you're referring to, > apart from handing the hardware devm allocated memory and unreferencing > the drm device? For udl if you unplug the physical device the struct device _will_ go away. The drm_device otoh might be referenced by piles of other internal things (like open file descriptors, shared dma-bufs and what else) and hence very much must stick around. Of course we /should/ try to not call down into lower-level hw handling code in that state. Which all very much doesn't happen (at least in a race-free manner). So the drm_device is not bounded by the lifetime of the underlying physical device in full generality. Which is why we can't really do any devm_* stuff in drm core. > > Not that anything in drm gets that right (pretty far from it), but > > that's why I haven't gone ballastic yet with rolling out devm_* all > > over our init code. > > I think you mean "ballistic", but, in this case, I think your typo > "ballastic" is actually a more accurate description of the change. :) Either way I think we should be ok with fully abusing devm_* ;-) -Daniel
diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 066a0ef..c069550 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1137,6 +1137,10 @@ int i915_driver_unload(struct drm_device *dev) dev_priv->vbt.child_dev = NULL; dev_priv->vbt.child_dev_num = 0; } + kfree(dev_priv->vbt.sdvo_lvds_vbt_mode); + dev_priv->vbt.sdvo_lvds_vbt_mode = NULL; + kfree(dev_priv->vbt.lfp_lvds_vbt_mode); + dev_priv->vbt.lfp_lvds_vbt_mode = NULL; vga_switcheroo_unregister_client(dev->pdev); vga_client_register(dev->pdev, NULL, NULL, NULL);
We allocate memory for LVDS modes while parsing the VBT at startup, but never free this memory when the driver is unloaded, causing a small leak. Signed-off-by: Matt Roper <matthew.d.roper@intel.com> --- drivers/gpu/drm/i915/i915_dma.c | 4 ++++ 1 file changed, 4 insertions(+)