Skip to contentSkip to conversation

mtd/nand: fix multi-chip suspend problem

PATCH In review [email protected] 3 participants 6 messages Nov 17, 2009 – Nov 26, 2009

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

PATCH 4/4 mtd/nand: fix multi-chip suspend problem

From: Li Yang <[email protected]>

Symptom:
device_suspend(): mtd_cls_suspend+0x0/0x58 returns -11
PM: Device mtd14 failed to suspend: error -11
PM: Some devices failed to suspend

This patch enables other chips to be suspended if the active chip of the controller has been suspended.

---

 drivers/mtd/nand/nand_base.c |   14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)
drivers/mtd/nand/nand_base.c renamed from drivers/mtd/nand/nand_base.c~mtd-nand-fix-multi-chip-suspend-problem +10 -4
diff -puN drivers/mtd/nand/nand_base.c~mtd-nand-fix-multi-chip-suspend-problem drivers/mtd/nand/nand_base.c
@@ -697,10 +697,16 @@ nand_get_device(struct nand_chip *chip, 
 		spin_unlock(lock);
 		return 0;
 	}
-	if (new_state == FL_PM_SUSPENDED) {
-		spin_unlock(lock);
-		return (chip->state == FL_PM_SUSPENDED) ? 0 : -EAGAIN;
-	}
+	if (new_state == FL_PM_SUSPENDED)
+		if (chip->controller->active->state == FL_PM_SUSPENDED) {
+			chip->state = FL_PM_SUSPENDED;
+			spin_unlock(lock);
+			return 0;
+		} else {
+			spin_unlock(lock);
+			return -EAGAIN;
+		}
+
 	set_current_state(TASK_UNINTERRUPTIBLE);
 	add_wait_queue(wq, &wait);
 	spin_unlock(lock);

_

#1 lore raw
Quoted 32 lines · On Tue, 2009-11-17 at 14:45 -0800, [email protected] wrote:

From: Li Yang <[email protected]>

Symptom:
device_suspend(): mtd_cls_suspend+0x0/0x58 returns -11
PM: Device mtd14 failed to suspend: error -11
PM: Some devices failed to suspend

This patch enables other chips to be suspended if the active chip of the controller has been suspended.

---

 drivers/mtd/nand/nand_base.c |   14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

diff -puN drivers/mtd/nand/nand_base.c~mtd-nand-fix-multi-chip-suspend-problem drivers/mtd/nand/nand_base.c

drivers/mtd/nand/nand_base.c renamed from drivers/mtd/nand/nand_base.c~mtd-nand-fix-multi-chip-suspend-problem +3 -4
@@ -697,10 +697,16 @@ nand_get_device(struct nand_chip *chip, 
 		spin_unlock(lock);
 		return 0;
 	}
-	if (new_state == FL_PM_SUSPENDED) {
-		spin_unlock(lock);
-		return (chip->state == FL_PM_SUSPENDED) ? 0 : -EAGAIN;
-	}
+	if (new_state == FL_PM_SUSPENDED)
+		if (chip->controller->active->state == FL_PM_SUSPENDED) {
+			chip->state = FL_PM_SUSPENDED;

drivers/mtd/nand/nand_base.c nand_get_device(struct nand_chip *chip, · line 712

Quoted 10 lines
drivers/mtd/nand/nand_base.c renamed from drivers/mtd/nand/nand_base.c~mtd-nand-fix-multi-chip-suspend-problem +7 -0
+			spin_unlock(lock);
+			return 0;
+		} else {
+			spin_unlock(lock);
+			return -EAGAIN;
+		}
+
 	set_current_state(TASK_UNINTERRUPTIBLE);
 	add_wait_queue(wq, &wait);
 	spin_unlock(lock);

This patch casing the following gcc warning:

drivers/mtd/nand/nand_base.c: In function ‘nand_get_device’:
drivers/mtd/nand/nand_base.c:700: warning: suggest explicit braces to avoid ambiguous ‘else’

Signature · 2 lines
Best Regards,
Artem Bityutskiy (Артём Битюцкий)
#2 lore raw

Ubiformat --erase-counter not always used

Hello Artem,

This is not a bug report, just an observation on ubiformat. It only seems to take the --erase-counter parameter in certain situations. If you have alien data it seems to take it. If the blocks are FF erased or previously formatted correctly it seems to ignore the --erase-counter argument, even though it says it is using it (see below).

Should we add something in the online help or documentation to explain this better and maybe say in the ubiformat tool itself when the switch is getting used or ignored and why?

# ./ubiformat /dev/mtd4 -n --erase-counter=1234 -v -y
ubiformat: mtd4 (NAND), size 2097152 bytes (2.0 MiB), 524288 eraseblocks of 524288 bytes (512.0 KiB), min. I/O size 4096 bytes
libscan: start scanning eraseblocks 0-4
libscan: scanning eraseblock 0: alien
libscan: scanning eraseblock 1: alien
libscan: scanning eraseblock 2: alien
libscan: scanning eraseblock 3: alien
libscan: finished, mean EC 0, 0 OK, 0 corrupted, 0 empty, 4 alien, bad 0
ubiformat: warning!: 4 of 4 eraseblocks contain non-ubifs data
ubiformat: use erase counter 1234 for all eraseblocks
ubiformat: eraseblock 0: erase, write EC 1234
ubiformat: eraseblock 1: erase, write EC 1234
ubiformat: eraseblock 2: erase, write EC 1234
ubiformat: eraseblock 3: erase, write EC 1234
192.168.1.105 # ./ubiformat /dev/mtd4 -n --erase-counter=5432 -v -y
ubiformat: mtd4 (NAND), size 2097152 bytes (2.0 MiB), 524288 eraseblocks of 524288 bytes (512.0 KiB), min. I/O size 4096 bytes
libscan: start scanning eraseblocks 0-4
libscan: scanning eraseblock 0: OK, erase counter 1234
libscan: scanning eraseblock 1: OK, erase counter 1234
libscan: scanning eraseblock 2: OK, erase counter 1234
libscan: scanning eraseblock 3: OK, erase counter 1234
libscan: finished, mean EC 1234, 4 OK, 0 corrupted, 0 empty, 0 alien, bad 0
ubiformat: 4 eraseblocks have valid erase counter, mean value is 1234
ubiformat: use erase counter 5432 for all eraseblocks    <------------------ Didn't take this value
ubiformat: eraseblock 0: erase, write EC 1235
ubiformat: eraseblock 1: erase, write EC 1235
ubiformat: eraseblock 2: erase, write EC 1235
ubiformat: eraseblock 3: erase, write EC 1235 

Thanks.

Regards,
Darwin

#3 lore raw

Re: Ubiformat --erase-counter not always used

Quoted 26 lines · Darwin Rambo

Hello Artem,

This is not a bug report, just an observation on ubiformat. It only seems to take the --erase-counter parameter in certain situations. If you have alien data it seems to take it. If the blocks are FF erased or previously formatted correctly it seems to ignore the --erase-counter argument, even though it says it is using it (see below).

Should we add something in the online help or documentation to explain this better and maybe say in the ubiformat tool itself when the switch is getting used or ignored and why?

# ./ubiformat /dev/mtd4 -n --erase-counter=1234 -v -y
ubiformat: mtd4 (NAND), size 2097152 bytes (2.0 MiB), 524288 eraseblocks of 524288 bytes (512.0 KiB), min. I/O size 4096 bytes
libscan: start scanning eraseblocks 0-4
libscan: scanning eraseblock 0: alien
libscan: scanning eraseblock 1: alien
libscan: scanning eraseblock 2: alien
libscan: scanning eraseblock 3: alien
libscan: finished, mean EC 0, 0 OK, 0 corrupted, 0 empty, 4 alien, bad 0
ubiformat: warning!: 4 of 4 eraseblocks contain non-ubifs data
ubiformat: use erase counter 1234 for all eraseblocks
ubiformat: eraseblock 0: erase, write EC 1234
ubiformat: eraseblock 1: erase, write EC 1234
ubiformat: eraseblock 2: erase, write EC 1234
ubiformat: eraseblock 3: erase, write EC 1234
192.168.1.105 # ./ubiformat /dev/mtd4 -n --erase-counter=5432 -v -y
ubiformat: mtd4 (NAND), size 2097152 bytes (2.0 MiB), 524288 eraseblocks of 524288 bytes (512.0 KiB), min. I/O size 4096 bytes
libscan: start scanning eraseblocks 0-4
libscan: scanning eraseblock 0: OK, erase counter 1234
libscan: scanning eraseblock 1: OK, erase counter 1234
libscan: scanning eraseblock 2: OK, erase counter 1234
Quoted 8 lines
libscan: scanning eraseblock 3: OK, erase counter 1234
libscan: finished, mean EC 1234, 4 OK, 0 corrupted, 0 empty, 0 alien, bad 0
ubiformat: 4 eraseblocks have valid erase counter, mean value is 1234
ubiformat: use erase counter 5432 for all eraseblocks    <------------------ Didn't take this value
ubiformat: eraseblock 0: erase, write EC 1235
ubiformat: eraseblock 1: erase, write EC 1235
ubiformat: eraseblock 2: erase, write EC 1235
ubiformat: eraseblock 3: erase, write EC 1235 

Looks like a bug to me. I think it should take the EC you ask it in all cases. I'll look at this when I have a bit more time, thanks for the report.

Signature · 2 lines
Best Regards,
Artem Bityutskiy (Артём Битюцкий)
#4 lore raw

Re: Ubiformat --erase-counter not always used

In reply to Darwin Rambo
Quoted 5 lines · Darwin Rambo

Hello Artem,

This is not a bug report, just an observation on ubiformat. It only seems to take the --erase-counter parameter in certain situations. If you have alien data it seems to take it. If the blocks are FF erased or previously formatted correctly it seems to ignore the --erase-counter argument, even though it says it is using it (see below).

Should we add something in the online help or documentation to explain this better and maybe say in the ubiformat tool itself when the switch is getting used or ignored and why?

Should be fixed in the mtd-utils repository now, thanks.

Signature · 2 lines
Best Regards,
Artem Bityutskiy (Артём Битюцкий)
#5 lore raw
In reply to Artem Bityutskiy
Quoted 46 lines · On Mon, 23 Nov 2009 17:19:04 +0200 Artem Bityutskiy <[email protected]> wrote:
Quoted 32 lines · On Tue, 2009-11-17 at 14:45 -0800, [email protected] wrote:

From: Li Yang <[email protected]>

Symptom:
device_suspend(): mtd_cls_suspend+0x0/0x58 returns -11
PM: Device mtd14 failed to suspend: error -11
PM: Some devices failed to suspend

This patch enables other chips to be suspended if the active chip of the controller has been suspended.

---

 drivers/mtd/nand/nand_base.c |   14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

diff -puN drivers/mtd/nand/nand_base.c~mtd-nand-fix-multi-chip-suspend-problem drivers/mtd/nand/nand_base.c

drivers/mtd/nand/nand_base.c renamed from drivers/mtd/nand/nand_base.c~mtd-nand-fix-multi-chip-suspend-problem +3 -4
@@ -697,10 +697,16 @@ nand_get_device(struct nand_chip *chip, 
 		spin_unlock(lock);
 		return 0;
 	}
-	if (new_state == FL_PM_SUSPENDED) {
-		spin_unlock(lock);
-		return (chip->state == FL_PM_SUSPENDED) ? 0 : -EAGAIN;
-	}
+	if (new_state == FL_PM_SUSPENDED)
+		if (chip->controller->active->state == FL_PM_SUSPENDED) {
+			chip->state = FL_PM_SUSPENDED;

drivers/mtd/nand/nand_base.c nand_get_device(struct nand_chip *chip, · line 712

Quoted 10 lines
drivers/mtd/nand/nand_base.c renamed from drivers/mtd/nand/nand_base.c~mtd-nand-fix-multi-chip-suspend-problem +7 -0
+			spin_unlock(lock);
+			return 0;
+		} else {
+			spin_unlock(lock);
+			return -EAGAIN;
+		}
+
 	set_current_state(TASK_UNINTERRUPTIBLE);
 	add_wait_queue(wq, &wait);
 	spin_unlock(lock);

This patch casing the following gcc warning:

Quoted 4 lines

drivers/mtd/nand/nand_base.c: In function ___nand_get_device___:
drivers/mtd/nand/nand_base.c:700: warning: suggest explicit braces to avoid ambiguous ___else___

Stupid gcc.

drivers/mtd/nand/nand_base.c renamed from drivers/mtd/nand/nand_base.c~mtd-nand-fix-multi-chip-suspend-problem-fix +2 -2
@@ -697,7 +697,7 @@ nand_get_device(struct nand_chip *chip, 
 		spin_unlock(lock);
 		return 0;
 	}
-	if (new_state == FL_PM_SUSPENDED)
+	if (new_state == FL_PM_SUSPENDED) {
 		if (chip->controller->active->state == FL_PM_SUSPENDED) {
 			chip->state = FL_PM_SUSPENDED;
 			spin_unlock(lock);
@@ -706,7 +706,7 @@ nand_get_device(struct nand_chip *chip, 
 			spin_unlock(lock);
 			return -EAGAIN;
 		}
-
+	}
 	set_current_state(TASK_UNINTERRUPTIBLE);
 	add_wait_queue(wq, &wait);
 	spin_unlock(lock);

_