Skip to contentSkip to conversation

tmio_mmc_pio: test TMIO_MMC_WRPROTECT_DISABLE earlier

PATCH In review Sergei Shtylyov 3 participants 7 messages Sep 23, 2015 – Sep 30, 2015

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tmio_mmc_pio: test TMIO_MMC_WRPROTECT_DISABLE earlier

There seems to be no sense in the runtime PM calls when the actual register read is suppressed by the TMIO_MMC_WRPROTECT_DISABLE flag. Check that flag before trying to read the register and thus doing the runtime PM dance...

While at it, kill useless local variable and add empty line after declarations.

---
The patch is against Ulf Hansson's 'mmc.git' repo's 'next' branch.

 drivers/mmc/host/tmio_mmc_pio.c |    8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
drivers/mmc/host/tmio_mmc_pio.c +5 -3
Index: mmc/drivers/mmc/host/tmio_mmc_pio.c
===================================================================
@@ -988,14 +988,16 @@ static void tmio_mmc_set_ios(struct mmc_
 static int tmio_mmc_get_ro(struct mmc_host *mmc)
 {
 	struct tmio_mmc_host *host = mmc_priv(mmc);
-	struct tmio_mmc_data *pdata = host->pdata;
 	int ret = mmc_gpio_get_ro(mmc);
+
 	if (ret >= 0)
 		return ret;
 
+	if (host->pdata->flags & TMIO_MMC_WRPROTECT_DISABLE)
+		return 0;
+
 	pm_runtime_get_sync(mmc_dev(mmc));
-	ret = !((pdata->flags & TMIO_MMC_WRPROTECT_DISABLE) ||
-		(sd_ctrl_read32(host, CTL_STATUS) & TMIO_STAT_WRPROTECT));
+	ret = !(sd_ctrl_read32(host, CTL_STATUS) & TMIO_STAT_WRPROTECT);
 	pm_runtime_mark_last_busy(mmc_dev(mmc));
 	pm_runtime_put_autosuspend(mmc_dev(mmc));
 
Acked-by on v1 1/1
#1 lore raw
Quoted 7 lines · Sergei Shtylyov

There seems to be no sense in the runtime PM calls when the actual register read is suppressed by the TMIO_MMC_WRPROTECT_DISABLE flag. Check that flag before trying to read the register and thus doing the runtime PM dance...

While at it, kill useless local variable and add empty line after declarations.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

#2 lore raw
In reply to Sergei Shtylyov
Quoted 27 lines · Sergei Shtylyov

There seems to be no sense in the runtime PM calls when the actual register read is suppressed by the TMIO_MMC_WRPROTECT_DISABLE flag. Check that flag before trying to read the register and thus doing the runtime PM dance...

While at it, kill useless local variable and add empty line after declarations.

---
The patch is against Ulf Hansson's 'mmc.git' repo's 'next' branch.

 drivers/mmc/host/tmio_mmc_pio.c |    8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

Index: mmc/drivers/mmc/host/tmio_mmc_pio.c =================================> --- mmc.orig/drivers/mmc/host/tmio_mmc_pio.c +++ mmc/drivers/mmc/host/tmio_mmc_pio.c

drivers/mmc/host/tmio_mmc_pio.c +1 -1
@@ -988,14 +988,16 @@ static void tmio_mmc_set_ios(struct mmc_
 static int tmio_mmc_get_ro(struct mmc_host *mmc)
 {
        struct tmio_mmc_host *host = mmc_priv(mmc);
-       struct tmio_mmc_data *pdata = host->pdata;
        int ret = mmc_gpio_get_ro(mmc);
+
        if (ret >= 0)
                return ret;
 

drivers/mmc/host/tmio_mmc_pio.c static void tmio_mmc_set_ios(struct mmc_ · line 1003

Quoted 10 lines
drivers/mmc/host/tmio_mmc_pio.c +4 -2
+       if (host->pdata->flags & TMIO_MMC_WRPROTECT_DISABLE)
+               return 0;
+
        pm_runtime_get_sync(mmc_dev(mmc));
-       ret = !((pdata->flags & TMIO_MMC_WRPROTECT_DISABLE) ||
-               (sd_ctrl_read32(host, CTL_STATUS) & TMIO_STAT_WRPROTECT));
+       ret = !(sd_ctrl_read32(host, CTL_STATUS) & TMIO_STAT_WRPROTECT);
        pm_runtime_mark_last_busy(mmc_dev(mmc));
        pm_runtime_put_autosuspend(mmc_dev(mmc));
 

Actually this change won't have the desired effect, as the mmc core already done a pm_runtime_get_sync() since it has claimed the mmc host[1].

I do realize that most drivers are still maintaining the pm_runtime_get|put() calls, but in most cases that's not needed any more.

[1]
commit 9250aea76bfc ("mmc: core: Enable runtime PM management of host devices")

Kind regards
Uffe

#3 lore raw

Hello.

Quoted 43 lines · Ulf Hansson
Quoted 28 lines

There seems to be no sense in the runtime PM calls when the actual register read is suppressed by the TMIO_MMC_WRPROTECT_DISABLE flag. Check that flag before trying to read the register and thus doing the runtime PM dance...

While at it, kill useless local variable and add empty line after declarations.

---
The patch is against Ulf Hansson's 'mmc.git' repo's 'next' branch.

  drivers/mmc/host/tmio_mmc_pio.c |    8 +++++---
  1 file changed, 5 insertions(+), 3 deletions(-)

Index: mmc/drivers/mmc/host/tmio_mmc_pio.c
===================================================================

drivers/mmc/host/tmio_mmc_pio.c +1 -1
@@ -988,14 +988,16 @@ static void tmio_mmc_set_ios(struct mmc_
  static int tmio_mmc_get_ro(struct mmc_host *mmc)
  {
         struct tmio_mmc_host *host = mmc_priv(mmc);
-       struct tmio_mmc_data *pdata = host->pdata;
         int ret = mmc_gpio_get_ro(mmc);
+
         if (ret >= 0)
                 return ret;
 

drivers/mmc/host/tmio_mmc_pio.c static void tmio_mmc_set_ios(struct mmc_ · line 1003

Quoted 10 lines
drivers/mmc/host/tmio_mmc_pio.c +4 -2
+       if (host->pdata->flags & TMIO_MMC_WRPROTECT_DISABLE)
+               return 0;
+
         pm_runtime_get_sync(mmc_dev(mmc));
-       ret = !((pdata->flags & TMIO_MMC_WRPROTECT_DISABLE) ||
-               (sd_ctrl_read32(host, CTL_STATUS) & TMIO_STAT_WRPROTECT));
+       ret = !(sd_ctrl_read32(host, CTL_STATUS) & TMIO_STAT_WRPROTECT);
         pm_runtime_mark_last_busy(mmc_dev(mmc));
         pm_runtime_put_autosuspend(mmc_dev(mmc));
 

Actually this change won't have the desired effect, as the mmc core already done a pm_runtime_get_sync() since it has claimed the mmc host[1].

Quoted 3 lines

I do realize that most drivers are still maintaining the pm_runtime_get|put() calls, but in most cases that's not needed any more.

Hm, OK. Should be OK to remove the RPM dances from at least this
particular driver, right?

Quoted 2 lines

[1]
commit 9250aea76bfc ("mmc: core: Enable runtime PM management of host devices")

Looking at the code, the following fragment of mmc_attach_sd() doesn't
make much sense to me:

        mmc_release_host(host);
        err = mmc_add_card(host->card);
        mmc_claim_host(host);
        if (err)
                goto remove_card;

        return 0;

remove_card:

        mmc_release_host(host);
        mmc_remove_card(host->card);
        host->card = NULL;
        mmc_claim_host(host);

Why claim the host and immediately release it on mmc_add_card() error? Can
we only claim on success and save a call here?

Quoted 2 lines

Kind regards
Uffe

MBR, Sergei

#4 lore raw

[...]

Quoted 6 lines
Quoted 4 lines

Actually this change won't have the desired effect, as the mmc core already done a pm_runtime_get_sync() since it has claimed the mmc host[1].

Quoted 7 lines
Quoted 3 lines

I do realize that most drivers are still maintaining the pm_runtime_get|put() calls, but in most cases that's not needed any more.

Hm, OK. Should be OK to remove the RPM dances from at least this
particular driver, right?

In most cases, yes.

There may be corner cases where the host is accessed without having the mmc core involved (and thus the host hasn't been claimed).

Quoted 17 lines
Quoted 4 lines

[1]
commit 9250aea76bfc ("mmc: core: Enable runtime PM management of host devices")

Looking at the code, the following fragment of mmc_attach_sd() doesn't
make much sense to me:

        mmc_release_host(host);
        err = mmc_add_card(host->card);
        mmc_claim_host(host);
        if (err)
                goto remove_card;

        return 0;
Quoted 8 lines

remove_card:

        mmc_release_host(host);
        mmc_remove_card(host->card);
        host->card = NULL;
        mmc_claim_host(host);

Why claim the host and immediately release it on mmc_add_card() error?
Can we only claim on success and save a call here?

You are right, we can simplify the sequence!

Do you want to send a patch? Actually, it's similar for mmc_attach_mmc()...

Kind regards
Uffe

#5 lore raw

Hello.

Quoted 10 lines · Ulf Hansson

[...]

Quoted 8 lines
Quoted 7 lines

Actually this change won't have the desired effect, as the mmc core already done a pm_runtime_get_sync() since it has claimed the mmc host[1].

I do realize that most drivers are still maintaining the pm_runtime_get|put() calls, but in most cases that's not needed any more.

Quoted 7 lines
Quoted 2 lines

Hm, OK. Should be OK to remove the RPM dances from at least this
particular driver, right?

In most cases, yes.

There may be corner cases where the host is accessed without having the mmc core involved (and thus the host hasn't been claimed).

OK, I'll try to look into this further...

Quoted 21 lines
Quoted 21 lines
Quoted 3 lines

[1]
commit 9250aea76bfc ("mmc: core: Enable runtime PM management of host devices")

Looking at the code, the following fragment of mmc_attach_sd() doesn't
make much sense to me:

         mmc_release_host(host);
         err = mmc_add_card(host->card);
         mmc_claim_host(host);
         if (err)
                 goto remove_card;

         return 0;

remove_card:

         mmc_release_host(host);
         mmc_remove_card(host->card);
         host->card = NULL;
         mmc_claim_host(host);
Quoted 4 lines
Quoted 2 lines

Why claim the host and immediately release it on mmc_add_card() error?
Can we only claim on success and save a call here?

You are right, we can simplify the sequence!

OK, what about calling mmc_remove_card() on mmc_add_card() failure?
Isn't it also superfluous?

Quoted 1 line

Do you want to send a patch?

Yes, I have it almost ready...

Quoted 1 line

Actually, it's similar for mmc_attach_mmc()...

Hm, will look into this case as well...

Quoted 2 lines

Kind regards
Uffe

MBR, Sergei

#6 lore raw

[...]

Quoted 23 lines
Quoted 21 lines
Quoted 21 lines

Looking at the code, the following fragment of mmc_attach_sd()
doesn't
make much sense to me:

         mmc_release_host(host);
         err = mmc_add_card(host->card);
         mmc_claim_host(host);
         if (err)
                 goto remove_card;

         return 0;

remove_card:

         mmc_release_host(host);
         mmc_remove_card(host->card);
         host->card = NULL;
         mmc_claim_host(host);

    Why claim the host and immediately release it on mmc_add_card()

error?
Can we only claim on success and save a call here?

Quoted 5 lines
Quoted 1 line

You are right, we can simplify the sequence!

OK, what about calling mmc_remove_card() on mmc_add_card() failure?
Isn't it also superfluous?

Nope.

Kind regards
Uffe