Message ID | 20200428215408.4111675-1-arnd@arndb.de |
---|---|
State | Accepted |
Commit | 78b0d99a68ecdc84728c99f4fef71942e9ecf35a |
Headers | show |
Series | drm/bridge: fix stack usage warning on old gcc | expand |
On 29/04/2020 00:53, Arnd Bergmann wrote: > Some older versions of gcc badly optimize code that passes > an inline function argument into another function by reference, > causing huge stack usage: > > drivers/gpu/drm/bridge/tc358768.c: In function 'tc358768_bridge_pre_enable': > drivers/gpu/drm/bridge/tc358768.c:840:1: error: the frame size of 2256 bytes is larger than 2048 bytes [-Werror=frame-larger-than=] > > Use a temporary variable as a workaround and add a comment pointing > to the gcc bug. > > Fixes: ff1ca6397b1d ("drm/bridge: Add tc358768 driver") > Signed-off-by: Arnd Bergmann <arnd@arndb.de> > --- > drivers/gpu/drm/bridge/tc358768.c | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > > diff --git a/drivers/gpu/drm/bridge/tc358768.c b/drivers/gpu/drm/bridge/tc358768.c > index 1b39e8d37834..6650fe4cfc20 100644 > --- a/drivers/gpu/drm/bridge/tc358768.c > +++ b/drivers/gpu/drm/bridge/tc358768.c > @@ -178,6 +178,8 @@ static int tc358768_clear_error(struct tc358768_priv *priv) > > static void tc358768_write(struct tc358768_priv *priv, u32 reg, u32 val) > { > + /* work around https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81715 */ > + int tmpval = val; > size_t count = 2; > > if (priv->error) > @@ -187,7 +189,7 @@ static void tc358768_write(struct tc358768_priv *priv, u32 reg, u32 val) > if (reg < 0x100 || reg >= 0x600) > count = 1; > > - priv->error = regmap_bulk_write(priv->regmap, reg, &val, count); > + priv->error = regmap_bulk_write(priv->regmap, reg, &tmpval, count); > } > > static void tc358768_read(struct tc358768_priv *priv, u32 reg, u32 *val) > tc358768_write is not inline. What is the inline function here? Or is tc358768_write optimized to inline by the compiler? Tomi -- Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
On Wed, Apr 29, 2020 at 9:56 AM Tomi Valkeinen <tomi.valkeinen@ti.com> wrote: > > diff --git a/drivers/gpu/drm/bridge/tc358768.c b/drivers/gpu/drm/bridge/tc358768.c > > index 1b39e8d37834..6650fe4cfc20 100644 > > --- a/drivers/gpu/drm/bridge/tc358768.c > > +++ b/drivers/gpu/drm/bridge/tc358768.c > > @@ -178,6 +178,8 @@ static int tc358768_clear_error(struct tc358768_priv *priv) > > > > static void tc358768_write(struct tc358768_priv *priv, u32 reg, u32 val) > > { > > + /* work around https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81715 */ > > + int tmpval = val; > > tc358768_write is not inline. What is the inline function here? Or is tc358768_write optimized to > inline by the compiler? I missed the lack of an explicit inline tag when looking at the bug. gcc usually decides which functions to inline on its own, so there is little difference in practice. Let me know if I should clarify the changelog and resend it. Arnd _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
On 29/04/2020 11:02, Arnd Bergmann wrote: > On Wed, Apr 29, 2020 at 9:56 AM Tomi Valkeinen <tomi.valkeinen@ti.com> wrote: >>> diff --git a/drivers/gpu/drm/bridge/tc358768.c b/drivers/gpu/drm/bridge/tc358768.c >>> index 1b39e8d37834..6650fe4cfc20 100644 >>> --- a/drivers/gpu/drm/bridge/tc358768.c >>> +++ b/drivers/gpu/drm/bridge/tc358768.c >>> @@ -178,6 +178,8 @@ static int tc358768_clear_error(struct tc358768_priv *priv) >>> >>> static void tc358768_write(struct tc358768_priv *priv, u32 reg, u32 val) >>> { >>> + /* work around https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81715 */ >>> + int tmpval = val; > >> >> tc358768_write is not inline. What is the inline function here? Or is tc358768_write optimized to >> inline by the compiler? > > I missed the lack of an explicit inline tag when looking at the bug. gcc > usually decides which functions to inline on its own, so there is little > difference in practice. Let me know if I should clarify the changelog and > resend it. Ok. I think this is fine. Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ti.com> Tomi -- Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
Hi Arnd. On Tue, Apr 28, 2020 at 11:53:54PM +0200, Arnd Bergmann wrote: > Some older versions of gcc badly optimize code that passes > an inline function argument into another function by reference, > causing huge stack usage: > > drivers/gpu/drm/bridge/tc358768.c: In function 'tc358768_bridge_pre_enable': > drivers/gpu/drm/bridge/tc358768.c:840:1: error: the frame size of 2256 bytes is larger than 2048 bytes [-Werror=frame-larger-than=] > > Use a temporary variable as a workaround and add a comment pointing > to the gcc bug. > > Fixes: ff1ca6397b1d ("drm/bridge: Add tc358768 driver") > Signed-off-by: Arnd Bergmann <arnd@arndb.de> Thanks, pushed to drm-misc-next with Tomi's r-b. Sam > --- > drivers/gpu/drm/bridge/tc358768.c | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > > diff --git a/drivers/gpu/drm/bridge/tc358768.c b/drivers/gpu/drm/bridge/tc358768.c > index 1b39e8d37834..6650fe4cfc20 100644 > --- a/drivers/gpu/drm/bridge/tc358768.c > +++ b/drivers/gpu/drm/bridge/tc358768.c > @@ -178,6 +178,8 @@ static int tc358768_clear_error(struct tc358768_priv *priv) > > static void tc358768_write(struct tc358768_priv *priv, u32 reg, u32 val) > { > + /* work around https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81715 */ > + int tmpval = val; > size_t count = 2; > > if (priv->error) > @@ -187,7 +189,7 @@ static void tc358768_write(struct tc358768_priv *priv, u32 reg, u32 val) > if (reg < 0x100 || reg >= 0x600) > count = 1; > > - priv->error = regmap_bulk_write(priv->regmap, reg, &val, count); > + priv->error = regmap_bulk_write(priv->regmap, reg, &tmpval, count); > } > > static void tc358768_read(struct tc358768_priv *priv, u32 reg, u32 *val) > -- > 2.26.0 > > _______________________________________________ > dri-devel mailing list > dri-devel@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/dri-devel _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
diff --git a/drivers/gpu/drm/bridge/tc358768.c b/drivers/gpu/drm/bridge/tc358768.c index 1b39e8d37834..6650fe4cfc20 100644 --- a/drivers/gpu/drm/bridge/tc358768.c +++ b/drivers/gpu/drm/bridge/tc358768.c @@ -178,6 +178,8 @@ static int tc358768_clear_error(struct tc358768_priv *priv) static void tc358768_write(struct tc358768_priv *priv, u32 reg, u32 val) { + /* work around https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81715 */ + int tmpval = val; size_t count = 2; if (priv->error) @@ -187,7 +189,7 @@ static void tc358768_write(struct tc358768_priv *priv, u32 reg, u32 val) if (reg < 0x100 || reg >= 0x600) count = 1; - priv->error = regmap_bulk_write(priv->regmap, reg, &val, count); + priv->error = regmap_bulk_write(priv->regmap, reg, &tmpval, count); } static void tc358768_read(struct tc358768_priv *priv, u32 reg, u32 *val)
Some older versions of gcc badly optimize code that passes an inline function argument into another function by reference, causing huge stack usage: drivers/gpu/drm/bridge/tc358768.c: In function 'tc358768_bridge_pre_enable': drivers/gpu/drm/bridge/tc358768.c:840:1: error: the frame size of 2256 bytes is larger than 2048 bytes [-Werror=frame-larger-than=] Use a temporary variable as a workaround and add a comment pointing to the gcc bug. Fixes: ff1ca6397b1d ("drm/bridge: Add tc358768 driver") Signed-off-by: Arnd Bergmann <arnd@arndb.de> --- drivers/gpu/drm/bridge/tc358768.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) -- 2.26.0 _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel