Skip to contentSkip to conversation

MMCI: use _cansleep GPIO functions

PATCH New Linus Walleij 5 participants 9 messages Sep 9, 2010 – Sep 11, 2010

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 MMCI: use _cansleep GPIO functions

Currently the kernel is screaming about slowpath at me for the wp/cd callbacks. Swicth to the _cansleep variants so as to silence this. Also remove the double-negate in front of the cd GPIO read.

---
 drivers/mmc/host/mmci.c |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)
drivers/mmc/host/mmci.c +3 -2
index f0c7313..082abad 100644
@@ -614,7 +614,7 @@ static int mmci_get_ro(struct mmc_host *mmc)
 	if (host->gpio_wp == -ENOSYS)
 		return -ENOSYS;
 
-	return gpio_get_value(host->gpio_wp);
+	return gpio_get_value_cansleep(host->gpio_wp);
 }
 
 static int mmci_get_cd(struct mmc_host *mmc)
@@ -629,7 +629,8 @@ static int mmci_get_cd(struct mmc_host *mmc)
 
 		status = plat->status(mmc_dev(host->mmc));
 	} else
-		status = !!gpio_get_value(host->gpio_cd) ^ plat->cd_invert;
+		status = gpio_get_value_cansleep(host->gpio_cd) ^
+			plat->cd_invert;
 
 	/*
 	 * Use positive logic throughout - status is zero for no card,
Signature · 1 line
1.6.3.3
#1 lore raw

PATCH 1/1 MMCI: use _cansleep GPIO functions

Quoted 3 lines · Linus Walleij

- ? ? ? ? ? ? ? status = !!gpio_get_value(host->gpio_cd) ^ plat->cd_invert;
+ ? ? ? ? ? ? ? status = gpio_get_value_cansleep(host->gpio_cd) ^
+ ? ? ? ? ? ? ? ? ? ? ? plat->cd_invert;

Why do you remove the double negation? gpio_get_value*() can return non-zero values, and we need to make it 1 for the inversion xor to work.

#2 lore raw

PATCH 1/1 MMCI: use _cansleep GPIO functions

2010/9/11 Rabin Vincent <[email protected]>:

Quoted 9 lines
Quoted 3 lines · Linus Walleij

- ? ? ? ? ? ? ? status = !!gpio_get_value(host->gpio_cd) ^ plat->cd_invert;
+ ? ? ? ? ? ? ? status = gpio_get_value_cansleep(host->gpio_cd) ^
+ ? ? ? ? ? ? ? ? ? ? ? plat->cd_invert;

Why do you remove the double negation? ?gpio_get_value*() can return non-zero values, and we need to make it 1 for the inversion xor to work.

Aha I get it. But !! is a bit hard to parse in head is it not...

Cast a non-zero to C internal bool type, negate it and then negate the negation so the end result is bool, not just non-zero. I never saw that idiom before.

I don't know if I'm particularly dumb but it makes my reading take time.

I'll slam in some intermediate bool or something so it's super clear what's happening, OK?

Yours,
Linus Walleij

#3 lore raw

PATCH 1/1 MMCI: use _cansleep GPIO functions

2010/9/11 Linus Walleij <[email protected]>:

Quoted 2 lines

I'll slam in some intermediate bool or something so it's super clear what's happening, OK?

Like this (untested):

From: Linus Walleij <[email protected]>
Date: Sat, 11 Sep 2010 13:29:04 +0200
Subject: [PATCH] MMCI: use _cansleep GPIO functions

Currently the kernel is screaming about slowpath at me for the wp/cd callbacks. Swicth to the _cansleep variants so as to silence this. Also rewrite the double-negate in front of the cd GPIO read to make it more clear what is happening.

---
 drivers/mmc/host/mmci.c |    6 ++++--
 1 files changed, 4 insertions(+), 2 deletions(-)
drivers/mmc/host/mmci.c +4 -2
index f0c7313..603d4d5 100644
@@ -614,7 +614,7 @@ static int mmci_get_ro(struct mmc_host *mmc)
 	if (host->gpio_wp == -ENOSYS)
 		return -ENOSYS;
 
-	return gpio_get_value(host->gpio_wp);
+	return gpio_get_value_cansleep(host->gpio_wp);
 }
 
 static int mmci_get_cd(struct mmc_host *mmc)
@@ -629,7 +629,9 @@ static int mmci_get_cd(struct mmc_host *mmc)
 
 		status = plat->status(mmc_dev(host->mmc));
 	} else
-		status = !!gpio_get_value(host->gpio_cd) ^ plat->cd_invert;
+		/* Convert the status to boolean and optionally invert */
+		status = (gpio_get_value_cansleep(host->gpio_cd) ? 1 : 0)
+			^ plat->cd_invert;
 
 	/*
 	 * Use positive logic throughout - status is zero for no card,
Signature · 1 line
1.7.2.2
#4 lore raw

PATCH 1/1 MMCI: use _cansleep GPIO functions

In reply to Linus Walleij
Quoted 2 lines · Linus Walleij

I'll slam in some intermediate bool or something so it's super clear what's happening, OK?

No - once you've learnt that !! turns an int into a bool, there's nothing more that needs to be done.

You'll find !! used throughout the kernel sources.

#5 lore raw

PATCH 1/1 MMCI: use _cansleep GPIO functions

2010/9/11 Russell King - ARM Linux <[email protected]>:

Quoted 1 line

You'll find !! used throughout the kernel sources.

Aha it's a kernel idiom, OK!!

I'll just test it on hardware and I'll put it in the patch tracker.

From: Linus Walleij <[email protected]>
Date: Sat, 11 Sep 2010 13:36:20 +0200
Subject: [PATCH] MMCI: use _cansleep GPIO functions

Currently the kernel is screaming about slowpath at me for the wp/cd callbacks. Swicth to the _cansleep variants so as to silence this.

---
 drivers/mmc/host/mmci.c |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)
drivers/mmc/host/mmci.c +3 -2
index f0c7313..c5fbf19 100644
@@ -614,7 +614,7 @@ static int mmci_get_ro(struct mmc_host *mmc)
 	if (host->gpio_wp == -ENOSYS)
 		return -ENOSYS;
 
-	return gpio_get_value(host->gpio_wp);
+	return gpio_get_value_cansleep(host->gpio_wp);
 }
 
 static int mmci_get_cd(struct mmc_host *mmc)
@@ -629,7 +629,8 @@ static int mmci_get_cd(struct mmc_host *mmc)
 
 		status = plat->status(mmc_dev(host->mmc));
 	} else
-		status = !!gpio_get_value(host->gpio_cd) ^ plat->cd_invert;
+		status = !!gpio_get_value_cansleep(host->gpio_cd)
+			^ plat->cd_invert;
 
 	/*
 	 * Use positive logic throughout - status is zero for no card,
Signature · 1 line
1.7.2.2
#6 lore raw

PATCH 1/1 MMCI: use _cansleep GPIO functions

Quoted 8 lines · Linus Walleij

2010/9/11 Russell King - ARM Linux <[email protected]>:

Quoted 1 line

You'll find !! used throughout the kernel sources.

Aha it's a kernel idiom, OK!!

I'll just test it on hardware and I'll put it in the patch tracker.

Quoted 7 lines

From: Linus Walleij <[email protected]>
Date: Sat, 11 Sep 2010 13:36:20 +0200
Subject: [PATCH] MMCI: use _cansleep GPIO functions

Currently the kernel is screaming about slowpath at me for the wp/cd callbacks. Swicth to the _cansleep variants so as to silence this.

You may wish to correct that spelling error.

#7 lore raw

PATCH 1/1 MMCI: use _cansleep GPIO functions

In reply to Linus Walleij

Hello.

Quoted 17 lines · Linus Walleij

From: Linus Walleij <[email protected]>
Date: Sat, 11 Sep 2010 13:36:20 +0200
Subject: [PATCH] MMCI: use _cansleep GPIO functions

Currently the kernel is screaming about slowpath at me for the wp/cd callbacks. Swicth to the _cansleep variants so as to silence this.

---
 drivers/mmc/host/mmci.c |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)
drivers/mmc/host/mmci.c +0 -0
index f0c7313..c5fbf19 100644

[...]

drivers/mmc/host/mmci.c static int mmci_get_cd(struct mmc_host *mmc) · line 633

Quoted 8 lines

static int mmci_get_cd(struct mmc_host *mmc)

drivers/mmc/host/mmci.c +2 -1
@@ -629,7 +629,8 @@ static int mmci_get_cd(struct mmc_host *mmc)
 
 		status = plat->status(mmc_dev(host->mmc));
 	} else
-		status = !!gpio_get_value(host->gpio_cd) ^ plat->cd_invert;
+		status = !!gpio_get_value_cansleep(host->gpio_cd)
+			^ plat->cd_invert;

Should be no space between ^ and operand, no?

WBR, Sergei

#8 lore raw

PATCH 1/1 MMCI: use _cansleep GPIO functions

Quoted 21 lines · I wrote:
Quoted 17 lines

From: Linus Walleij <[email protected]>
Date: Sat, 11 Sep 2010 13:36:20 +0200
Subject: [PATCH] MMCI: use _cansleep GPIO functions

Currently the kernel is screaming about slowpath at me for the wp/cd callbacks. Swicth to the _cansleep variants so as to silence this.

---
 drivers/mmc/host/mmci.c |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)
drivers/mmc/host/mmci.c +0 -0
index f0c7313..c5fbf19 100644

[...]

Quoted 3 lines

static int mmci_get_cd(struct mmc_host *mmc)

drivers/mmc/host/mmci.c +0 -0
@@ -629,7 +629,8 @@ static int mmci_get_cd(struct mmc_host *mmc)
 
Quoted 7 lines

drivers/mmc/host/mmci.c static int mmci_get_cd(struct mmc_host *mmc) · line 633

Quoted 5 lines
drivers/mmc/host/mmci.c +2 -1
         status = plat->status(mmc_dev(host->mmc));
     } else
-        status = !!gpio_get_value(host->gpio_cd) ^ plat->cd_invert;
+        status = !!gpio_get_value_cansleep(host->gpio_cd)
+            ^ plat->cd_invert;

Should be no space between ^ and operand, no?

Oops, I've mixed it with ~. Still, I'd have left ^ on the first line.

WBR, Sergei