Skip to contentSkip to conversation

tpm: fix compat 'ppi' link handling in tpm_chip_register()

PATCH In review Jarkko Sakkinen 3 participants 5 messages Nov 8, 2015 – Nov 10, 2015 TPM device driver

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tpm: fix compat 'ppi' link handling in tpm_chip_register()

__compat_only_sysfs_link_entry_to_kobj() was unconditionally called for TPM1 chips, which caused crash on Acer C720 laptop where DSM for the ACPI object did not exist.

There are two reasons for unwanted behavior:

* The code did not check whether __compat_only_sysfs_link_entry_to_kobj() returned -ENOENT. This is OK. It just meanst that ppi is not available.
* The code did not clean up properly. Compat link should added only after all other init is done.

This patch sorts out these issues.

---
 drivers/char/tpm/tpm-chip.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)
drivers/char/tpm/tpm-chip.c +10 -8
index a5cdce7..45cc39a 100644
@@ -226,14 +226,6 @@ int tpm_chip_register(struct tpm_chip *chip)
 	if (rc)
 		goto out_err;
 
-	if (!(chip->flags & TPM_CHIP_FLAG_TPM2)) {
-		rc = __compat_only_sysfs_link_entry_to_kobj(&chip->pdev->kobj,
-							    &chip->dev.kobj,
-							    "ppi");
-		if (rc)
-			goto out_err;
-	}
-
 	/* Make the chip available. */
 	spin_lock(&driver_lock);
 	list_add_tail_rcu(&chip->list, &tpm_chip_list);
@@ -241,6 +233,16 @@ int tpm_chip_register(struct tpm_chip *chip)
 
 	chip->flags |= TPM_CHIP_FLAG_REGISTERED;
 
+	if (!(chip->flags & TPM_CHIP_FLAG_TPM2)) {
+		rc = __compat_only_sysfs_link_entry_to_kobj(&chip->pdev->kobj,
+							    &chip->dev.kobj,
+							    "ppi");
+		if (rc && rc != -ENOENT) {
+			tpm_chip_unregister(chip);
+			return rc;
+		}
+	}
+
 	return 0;
 out_err:
 	tpm1_chip_unregister(chip);
Signature · 1 line
2.5.0
#1 lore raw

drivers/char/tpm/tpm-chip.c int tpm_chip_register(struct tpm_chip *chip) · line 244

Quoted 5 lines · Jarkko Sakkinen
drivers/char/tpm/tpm-chip.c +5 -0
+		if (rc && rc != -ENOENT) {
+			tpm_chip_unregister(chip);
+			return rc;
+		}
+	}

This is still goofy looking, the list_add_tail_rcu should be the last thing done and cannot fail. Just code this with the usual goto based unwind and goto to do tpm_dev_del_device for the above failure.

drivers/char/tpm/tpm-chip.c int tpm_chip_register(struct tpm_chip *chip) · line 248

Quoted 4 lines
drivers/char/tpm/tpm-chip.c +1 -0
+
 	return 0;
 out_err:
 	tpm1_chip_unregister(chip);

Which avoids calling tpm1_chip_unregister and tpm_chip_unregister in the same function, which looks so wrong to a casual read..

Jason

#2 lore raw
Quoted 10 lines · Jason Gunthorpe

drivers/char/tpm/tpm-chip.c int tpm_chip_register(struct tpm_chip *chip) · line 244

Quoted 5 lines · Jarkko Sakkinen
drivers/char/tpm/tpm-chip.c +5 -0
+		if (rc && rc != -ENOENT) {
+			tpm_chip_unregister(chip);
+			return rc;
+		}
+	}

This is still goofy looking, the list_add_tail_rcu should be the last thing done and cannot fail. Just code this with the usual goto based unwind and goto to do tpm_dev_del_device for the above failure.

Agreed that this would be cleaner.

Quoted 7 lines

drivers/char/tpm/tpm-chip.c int tpm_chip_register(struct tpm_chip *chip) · line 248

Quoted 4 lines
drivers/char/tpm/tpm-chip.c +1 -0
+
 	return 0;
 out_err:
 	tpm1_chip_unregister(chip);

Which avoids calling tpm1_chip_unregister and tpm_chip_unregister in the same function, which looks so wrong to a casual read..

I'm not going to change the current patch at this point of release cycle but this clean up definitely makes sense later on.

Quoted 1 line

Jason

/Jarkko

#3 lore raw
In reply to Jarkko Sakkinen

Jarkko,

Quoted 61 lines · Jarkko Sakkinen

__compat_only_sysfs_link_entry_to_kobj() was unconditionally called for TPM1 chips, which caused crash on Acer C720 laptop where DSM for the ACPI object did not exist.

There are two reasons for unwanted behavior:

* The code did not check whether __compat_only_sysfs_link_entry_to_kobj() returned -ENOENT. This is OK. It just meanst that ppi is not available.
* The code did not clean up properly. Compat link should added only after all other init is done.

This patch sorts out these issues.

---
 drivers/char/tpm/tpm-chip.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)
drivers/char/tpm/tpm-chip.c +10 -8
index a5cdce7..45cc39a 100644
@@ -226,14 +226,6 @@ int tpm_chip_register(struct tpm_chip *chip)
 	if (rc)
 		goto out_err;
 
-	if (!(chip->flags & TPM_CHIP_FLAG_TPM2)) {
-		rc = __compat_only_sysfs_link_entry_to_kobj(&chip->pdev->kobj,
-							    &chip->dev.kobj,
-							    "ppi");
-		if (rc)
-			goto out_err;
-	}
-
 	/* Make the chip available. */
 	spin_lock(&driver_lock);
 	list_add_tail_rcu(&chip->list, &tpm_chip_list);
@@ -241,6 +233,16 @@ int tpm_chip_register(struct tpm_chip *chip)
 
 	chip->flags |= TPM_CHIP_FLAG_REGISTERED;
 
+	if (!(chip->flags & TPM_CHIP_FLAG_TPM2)) {
+		rc = __compat_only_sysfs_link_entry_to_kobj(&chip->pdev->kobj,
+							    &chip->dev.kobj,
+							    "ppi");
+		if (rc && rc != -ENOENT) {
+			tpm_chip_unregister(chip);
+			return rc;
+		}
+	}
+
 	return 0;
 out_err:
 	tpm1_chip_unregister(chip);
Signature · 1 line
2.5.0

This patch doesn't apply to the latest linux-next (20151109). I think I may be missing a dependent patch. Do you know which one?

Signature · 1 line
- Jeremiah Mahler
#4 lore raw
Quoted 59 lines · Jeremiah Mahler

Jarkko,

Quoted 56 lines · Jarkko Sakkinen

__compat_only_sysfs_link_entry_to_kobj() was unconditionally called for TPM1 chips, which caused crash on Acer C720 laptop where DSM for the ACPI object did not exist.

There are two reasons for unwanted behavior:

* The code did not check whether __compat_only_sysfs_link_entry_to_kobj() returned -ENOENT. This is OK. It just meanst that ppi is not available.
* The code did not clean up properly. Compat link should added only after all other init is done.

This patch sorts out these issues.

---
 drivers/char/tpm/tpm-chip.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)
drivers/char/tpm/tpm-chip.c +10 -8
index a5cdce7..45cc39a 100644
@@ -226,14 +226,6 @@ int tpm_chip_register(struct tpm_chip *chip)
 	if (rc)
 		goto out_err;
 
-	if (!(chip->flags & TPM_CHIP_FLAG_TPM2)) {
-		rc = __compat_only_sysfs_link_entry_to_kobj(&chip->pdev->kobj,
-							    &chip->dev.kobj,
-							    "ppi");
-		if (rc)
-			goto out_err;
-	}
-
 	/* Make the chip available. */
 	spin_lock(&driver_lock);
 	list_add_tail_rcu(&chip->list, &tpm_chip_list);
@@ -241,6 +233,16 @@ int tpm_chip_register(struct tpm_chip *chip)
 
 	chip->flags |= TPM_CHIP_FLAG_REGISTERED;
 
+	if (!(chip->flags & TPM_CHIP_FLAG_TPM2)) {
+		rc = __compat_only_sysfs_link_entry_to_kobj(&chip->pdev->kobj,
+							    &chip->dev.kobj,
+							    "ppi");
+		if (rc && rc != -ENOENT) {
+			tpm_chip_unregister(chip);
+			return rc;
+		}
+	}
+
 	return 0;
Quoted 8 lines
Quoted 5 lines
drivers/char/tpm/tpm-chip.c +0 -0
 out_err:
 	tpm1_chip_unregister(chip);
Signature · 1 line
2.5.0

This patch doesn't apply to the latest linux-next (20151109). I think I may be missing a dependent patch. Do you know which one?

It is expected as there are 7 other bug fixes that I'm including to the next pull rquest and this is on top.

Quoted 1 line

- Jeremiah Mahler

/Jarkko