| Message ID | fc824f91-f686-64ab-83e4-86b04fd512f0@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Heiner Kallweit <hkallweit1@gmail.com> writes: > This clk_disable_unprepare isn't right here. At first the condition > should be "if (ret)" as the disable/unprepare is supposed to be > executed if the previous command fails. > And if the previous command fails and meson_mmc_clk_init returns > an error, then probe jumps to label free_host where we have this > disable/unprepare already. > > Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> Acked-by: Kevin Hilman <khilman@baylibre.com> > --- > v2: > - extended commit message > --- > drivers/mmc/host/meson-gx-mmc.c | 6 +----- > 1 file changed, 1 insertion(+), 5 deletions(-) > > diff --git a/drivers/mmc/host/meson-gx-mmc.c b/drivers/mmc/host/meson-gx-mmc.c > index 68e76fa8..850b0152 100644 > --- a/drivers/mmc/host/meson-gx-mmc.c > +++ b/drivers/mmc/host/meson-gx-mmc.c > @@ -320,11 +320,7 @@ static int meson_mmc_clk_init(struct meson_host *host) > /* Get the nearest minimum clock to 400KHz */ > host->mmc->f_min = clk_round_rate(host->cfg_div_clk, 400000); > > - ret = meson_mmc_clk_set(host, host->mmc->f_min); > - if (!ret) > - clk_disable_unprepare(host->cfg_div_clk); > - > - return ret; > + return meson_mmc_clk_set(host, host->mmc->f_min); > } > > static void meson_mmc_set_ios(struct mmc_host *mmc, struct mmc_ios *ios) -- To unsubscribe from this list: send the line "unsubscribe linux-mmc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hello Heiner, W dniu 16.02.2017 o 20:42, Heiner Kallweit pisze: > This clk_disable_unprepare isn't right here. At first the condition > should be "if (ret)" as the disable/unprepare is supposed to be > executed if the previous command fails. > And if the previous command fails and meson_mmc_clk_init returns > an error, then probe jumps to label free_host where we have this > disable/unprepare already. > > Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> > --- > v2: > - extended commit message > --- > drivers/mmc/host/meson-gx-mmc.c | 6 +----- > 1 file changed, 1 insertion(+), 5 deletions(-) > > diff --git a/drivers/mmc/host/meson-gx-mmc.c b/drivers/mmc/host/meson-gx-mmc.c > index 68e76fa8..850b0152 100644 > --- a/drivers/mmc/host/meson-gx-mmc.c > +++ b/drivers/mmc/host/meson-gx-mmc.c > @@ -320,11 +320,7 @@ static int meson_mmc_clk_init(struct meson_host *host) > /* Get the nearest minimum clock to 400KHz */ > host->mmc->f_min = clk_round_rate(host->cfg_div_clk, 400000); > > - ret = meson_mmc_clk_set(host, host->mmc->f_min); > - if (!ret) > - clk_disable_unprepare(host->cfg_div_clk); > - > - return ret; > + return meson_mmc_clk_set(host, host->mmc->f_min); > } > > static void meson_mmc_set_ios(struct mmc_host *mmc, struct mmc_ios *ios) > I see one problem here. As you said, if the previous command (that is the meson_mmc_clk_set function call) fails, then the whole meson_mmc_clk_init returns an error value, and meson_mmc_probe jumps to free_host and disables the clock there. But what happens if the clk_prepare_enable is the line that fails? in this case, the function returns an error value too, and in the free_host meson_mmc_probe tries to disable the clock that is not enabled, and that probably results in a kernel warning because you tried to disable a disabled clock. I was preparing a patch to disable clocks in the probe function only if they were enabled before, but then the meson_mmc_clk_init function remains a problem, because we don't know if the clock was enabled successfully or not when it returns an error.
Am 18.02.2017 um 11:37 schrieb Heiner Kallweit: > Am 18.02.2017 um 04:44 schrieb Michał Zegan: >> Hello Heiner, >> >> W dniu 16.02.2017 o 20:42, Heiner Kallweit pisze: >>> This clk_disable_unprepare isn't right here. At first the condition >>> should be "if (ret)" as the disable/unprepare is supposed to be >>> executed if the previous command fails. >>> And if the previous command fails and meson_mmc_clk_init returns >>> an error, then probe jumps to label free_host where we have this >>> disable/unprepare already. >>> >>> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> >>> --- >>> v2: >>> - extended commit message >>> --- >>> drivers/mmc/host/meson-gx-mmc.c | 6 +----- >>> 1 file changed, 1 insertion(+), 5 deletions(-) >>> >>> diff --git a/drivers/mmc/host/meson-gx-mmc.c b/drivers/mmc/host/meson-gx-mmc.c >>> index 68e76fa8..850b0152 100644 >>> --- a/drivers/mmc/host/meson-gx-mmc.c >>> +++ b/drivers/mmc/host/meson-gx-mmc.c >>> @@ -320,11 +320,7 @@ static int meson_mmc_clk_init(struct meson_host *host) >>> /* Get the nearest minimum clock to 400KHz */ >>> host->mmc->f_min = clk_round_rate(host->cfg_div_clk, 400000); >>> >>> - ret = meson_mmc_clk_set(host, host->mmc->f_min); >>> - if (!ret) >>> - clk_disable_unprepare(host->cfg_div_clk); >>> - >>> - return ret; >>> + return meson_mmc_clk_set(host, host->mmc->f_min); >>> } >>> >>> static void meson_mmc_set_ios(struct mmc_host *mmc, struct mmc_ios *ios) >>> >> >> I see one problem here. >> As you said, if the previous command (that is the meson_mmc_clk_set >> function call) fails, then the whole meson_mmc_clk_init returns an error >> value, and meson_mmc_probe jumps to free_host and disables the clock there. >> But what happens if the clk_prepare_enable is the line that fails? in >> this case, the function returns an error value too, and in the free_host >> meson_mmc_probe tries to disable the clock that is not enabled, and that >> probably results in a kernel warning because you tried to disable a >> disabled clock. >> I was preparing a patch to disable clocks in the probe function only if >> they were enabled before, but then the meson_mmc_clk_init function >> remains a problem, because we don't know if the clock was enabled >> successfully or not when it returns an error. >> > Thanks for pointing out, you're right. The solution however should be > quite easy. I meson_mmc_clk_init we just change the condition of > the disable/unprepare to "if (ret)" and in probe we remove this > disable/unprepare from the error path. > This was a little too fast, it's not that easy. But I'll come up with a changed version. -- To unsubscribe from this list: send the line "unsubscribe linux-mmc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
W dniu 18.02.2017 o 11:37, Heiner Kallweit pisze: > Am 18.02.2017 um 04:44 schrieb Michał Zegan: >> Hello Heiner, >> >> W dniu 16.02.2017 o 20:42, Heiner Kallweit pisze: >>> This clk_disable_unprepare isn't right here. At first the condition >>> should be "if (ret)" as the disable/unprepare is supposed to be >>> executed if the previous command fails. >>> And if the previous command fails and meson_mmc_clk_init returns >>> an error, then probe jumps to label free_host where we have this >>> disable/unprepare already. >>> >>> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> >>> --- >>> v2: >>> - extended commit message >>> --- >>> drivers/mmc/host/meson-gx-mmc.c | 6 +----- >>> 1 file changed, 1 insertion(+), 5 deletions(-) >>> >>> diff --git a/drivers/mmc/host/meson-gx-mmc.c b/drivers/mmc/host/meson-gx-mmc.c >>> index 68e76fa8..850b0152 100644 >>> --- a/drivers/mmc/host/meson-gx-mmc.c >>> +++ b/drivers/mmc/host/meson-gx-mmc.c >>> @@ -320,11 +320,7 @@ static int meson_mmc_clk_init(struct meson_host *host) >>> /* Get the nearest minimum clock to 400KHz */ >>> host->mmc->f_min = clk_round_rate(host->cfg_div_clk, 400000); >>> >>> - ret = meson_mmc_clk_set(host, host->mmc->f_min); >>> - if (!ret) >>> - clk_disable_unprepare(host->cfg_div_clk); >>> - >>> - return ret; >>> + return meson_mmc_clk_set(host, host->mmc->f_min); >>> } >>> >>> static void meson_mmc_set_ios(struct mmc_host *mmc, struct mmc_ios *ios) >>> >> >> I see one problem here. >> As you said, if the previous command (that is the meson_mmc_clk_set >> function call) fails, then the whole meson_mmc_clk_init returns an error >> value, and meson_mmc_probe jumps to free_host and disables the clock there. >> But what happens if the clk_prepare_enable is the line that fails? in >> this case, the function returns an error value too, and in the free_host >> meson_mmc_probe tries to disable the clock that is not enabled, and that >> probably results in a kernel warning because you tried to disable a >> disabled clock. >> I was preparing a patch to disable clocks in the probe function only if >> they were enabled before, but then the meson_mmc_clk_init function >> remains a problem, because we don't know if the clock was enabled >> successfully or not when it returns an error. >> > Thanks for pointing out, you're right. The solution however should be > quite easy. I meson_mmc_clk_init we just change the condition of > the disable/unprepare to "if (ret)" and in probe we remove this > disable/unprepare from the error path. > If you do that, then if something fails *after* the meson_mmc_clk_init returns success, but something else fails later, then you will end up not disabling those clocks. What I thought about is: change the condition as you said. but in meson_mmc_probe, either check if clock is enabled in free_host before disabling it, or have two additional error passes: one used after the core_clk is enabled (currently if it fails then it will also be a kernel warning), and one used after the meson_mmc_clk_init returns success that would also disable the div clock. Not sure if that second one is appropriate or not. > > _______________________________________________ > linux-amlogic mailing list > linux-amlogic@lists.infradead.org > http://lists.infradead.org/mailman/listinfo/linux-amlogic >
diff --git a/drivers/mmc/host/meson-gx-mmc.c b/drivers/mmc/host/meson-gx-mmc.c index 68e76fa8..850b0152 100644 --- a/drivers/mmc/host/meson-gx-mmc.c +++ b/drivers/mmc/host/meson-gx-mmc.c @@ -320,11 +320,7 @@ static int meson_mmc_clk_init(struct meson_host *host) /* Get the nearest minimum clock to 400KHz */ host->mmc->f_min = clk_round_rate(host->cfg_div_clk, 400000); - ret = meson_mmc_clk_set(host, host->mmc->f_min); - if (!ret) - clk_disable_unprepare(host->cfg_div_clk); - - return ret; + return meson_mmc_clk_set(host, host->mmc->f_min); } static void meson_mmc_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)
This clk_disable_unprepare isn't right here. At first the condition should be "if (ret)" as the disable/unprepare is supposed to be executed if the previous command fails. And if the previous command fails and meson_mmc_clk_init returns an error, then probe jumps to label free_host where we have this disable/unprepare already. Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> --- v2: - extended commit message --- drivers/mmc/host/meson-gx-mmc.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-)