Skip to contentSkip to conversation

tpm: Remove dead NULL check in tpm2_flush_space()

PATCH In review Gunnar Kudrjavets 3 participants 9 messages Apr 27 – May 10

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

PATCH 1/1 tpm: Remove dead NULL check in tpm2_flush_space()

The 'space' pointer in tpm2_flush_space() is assigned from &chip->work_space, which is the address of an embedded struct member within struct tpm_chip. This address can never be NULL, making the NULL check dead code. The new code follows the existing pattern established by the other callers in tpm2-space.c which also assign from &chip->work_space without a NULL check. Remove the dead code to avoid confusion.

Fixes e3aaebcbb7c6 ("tpm: Clean up TPM space after command failure")
Assisted-by Kiro:claude-opus-4.6
---
 drivers/char/tpm/tpm2-space.c | 3 ---
 1 file changed, 3 deletions(-)
drivers/char/tpm/tpm2-space.c +0 -3
index 60354cd53b5c..1eec72eb8208 100644
@@ -169,9 +169,6 @@ void tpm2_flush_space(struct tpm_chip *chip)
 	struct tpm_space *space = &chip->work_space;
 	int i;
 
-	if (!space)
-		return;
-
 	for (i = 0; i < ARRAY_SIZE(space->context_tbl); i++)
 		if (space->context_tbl[i] && ~space->context_tbl[i])
 			tpm2_flush_context(chip, space->context_tbl[i]);

base-commit: 949692da7211572fac419b2986b6abc0cd1aeb76

Signature · 1 line
2.47.3
#1 lore raw

Dear Gunnar,

Thank you for your patch.

Am 27.04.26 um 18:32 schrieb Gunnar Kudrjavets:

Quoted 21 lines

The 'space' pointer in tpm2_flush_space() is assigned from &chip->work_space, which is the address of an embedded struct member within struct tpm_chip. This address can never be NULL, making the NULL check dead code. The new code follows the existing pattern established by the other callers in tpm2-space.c which also assign from &chip->work_space without a NULL check. Remove the dead code to avoid confusion.

Fixes e3aaebcbb7c6 ("tpm: Clean up TPM space after command failure")
Assisted-by Kiro:claude-opus-4.6
---
  drivers/char/tpm/tpm2-space.c | 3 ---
  1 file changed, 3 deletions(-)
drivers/char/tpm/tpm2-space.c +0 -0
index 60354cd53b5c..1eec72eb8208 100644

drivers/char/tpm/tpm2-space.c void tpm2_flush_space(struct tpm_chip *chip) · line 174

Quoted 9 lines
drivers/char/tpm/tpm2-space.c +0 -3
@@ -169,9 +169,6 @@ void tpm2_flush_space(struct tpm_chip *chip)
  	struct tpm_space *space = &chip->work_space;
  	int i;
 
-	if (!space)
-		return;
-
  	for (i = 0; i < ARRAY_SIZE(space->context_tbl); i++)
  		if (space->context_tbl[i] && ~space->context_tbl[i])
  			tpm2_flush_context(chip, space->context_tbl[i]);

gemini/gemini-3.1-pro-preview made a comment [1]. No idea, if it’s valid.

Kind regards,

Paul

[1]: https://sashiko.dev/#/patchset/20260427163238.20230-1-gunnarku%40amazon.com

#2 lore raw
Quoted 1 line · Paul Menzel

gemini/gemini-3.1-pro-preview made a comment [1]. No idea, if it's valid.

Thanks for forwarding, Paul. AFAICS, the comment is a false positive.

My theory is that Gemini conflates two different variables named 'space':

1. The 'space' parameter passed to tpm_dev_transmit(). This *can* be NULL (it is NULL for /dev/tpm0 clients).

2. The local 'space' variable inside tpm2_flush_space(). This is assigned from &chip->work_space and can *never* be NULL.

The removed NULL check was testing case (2), not case (1).

Regards,
Gunnar

#3 lore raw

Dear Gunnar,

Am 28.04.26 um 00:57 schrieb Gunnar Kudrjavets:

Quoted 8 lines · Paul Menzel
Quoted 1 line

gemini/gemini-3.1-pro-preview made a comment [1]. No idea, if it's valid.

Thanks for forwarding, Paul. AFAICS, the comment is a false positive.

My theory is that Gemini conflates two different variables named 'space':

Quoted 7 lines

1. The 'space' parameter passed to tpm_dev_transmit(). This *can* be NULL (it is NULL for /dev/tpm0 clients).

2. The local 'space' variable inside tpm2_flush_space(). This is

    assigned from &chip->work_space and can *never* be NULL.

The removed NULL check was testing case (2), not case (1).

Thank you for quickly looking into this. Sorry for the noise.

Kind regards,

Paul

maintainer
#4 lore raw
Quoted 15 lines · Paul Menzel

Dear Gunnar,

Am 28.04.26 um 00:57 schrieb Gunnar Kudrjavets:

Quoted 11 lines · Paul Menzel
Quoted 1 line

gemini/gemini-3.1-pro-preview made a comment [1]. No idea, if it's valid.

Thanks for forwarding, Paul. AFAICS, the comment is a false positive.

My theory is that Gemini conflates two different variables named 'space':

1. The 'space' parameter passed to tpm_dev_transmit(). This *can* be NULL (it is NULL for /dev/tpm0 clients).

Quoted 6 lines
Quoted 4 lines

2. The local 'space' variable inside tpm2_flush_space(). This is

    assigned from &chip->work_space and can *never* be NULL.

The removed NULL check was testing case (2), not case (1).

Thank you for quickly looking into this. Sorry for the noise.

I felt sleep while reading the text generated by Gemini but yes exactly because it can't be NULL the check is there.

I.e. Opus is right but the action taken is still plain wrong.

Quoted 3 lines

Kind regards,

Paul

BR, Jarkko

maintainer
#5 lore raw
In reply to Gunnar Kudrjavets
Quoted 11 lines · Gunnar Kudrjavets

The 'space' pointer in tpm2_flush_space() is assigned from &chip->work_space, which is the address of an embedded struct member within struct tpm_chip. This address can never be NULL, making the NULL check dead code. The new code follows the existing pattern established by the other callers in tpm2-space.c which also assign from &chip->work_space without a NULL check. Remove the dead code to avoid confusion.

Fixes e3aaebcbb7c6 ("tpm: Clean up TPM space after command failure")
Assisted-by Kiro:claude-opus-4.6

Just for sake of understanding:

What is "kiro" and is assisted-by the tag supposed to be used here?

Quoted 20 lines
---
 drivers/char/tpm/tpm2-space.c | 3 ---
 1 file changed, 3 deletions(-)
drivers/char/tpm/tpm2-space.c +0 -3
index 60354cd53b5c..1eec72eb8208 100644
@@ -169,9 +169,6 @@ void tpm2_flush_space(struct tpm_chip *chip)
 	struct tpm_space *space = &chip->work_space;
 	int i;
 
-	if (!space)
-		return;
-
 	for (i = 0; i < ARRAY_SIZE(space->context_tbl); i++)
 		if (space->context_tbl[i] && ~space->context_tbl[i])
 			tpm2_flush_context(chip, space->context_tbl[i]);
Quoted 4 lines

base-commit: 949692da7211572fac419b2986b6abc0cd1aeb76

Signature · 1 line
2.47.3

It's all good otherwise, just need clarification as we are learning how to deal with these patches :-)

BR, Jarkko

maintainer
#6 lore raw
Quoted 37 lines · Jarkko Sakkinen
Quoted 11 lines · Gunnar Kudrjavets

The 'space' pointer in tpm2_flush_space() is assigned from &chip->work_space, which is the address of an embedded struct member within struct tpm_chip. This address can never be NULL, making the NULL check dead code. The new code follows the existing pattern established by the other callers in tpm2-space.c which also assign from &chip->work_space without a NULL check. Remove the dead code to avoid confusion.

Fixes e3aaebcbb7c6 ("tpm: Clean up TPM space after command failure")
Assisted-by Kiro:claude-opus-4.6

Just for sake of understanding:

What is "kiro" and is assisted-by the tag supposed to be used here?

Quoted 20 lines
---
 drivers/char/tpm/tpm2-space.c | 3 ---
 1 file changed, 3 deletions(-)
drivers/char/tpm/tpm2-space.c +0 -3
index 60354cd53b5c..1eec72eb8208 100644
@@ -169,9 +169,6 @@ void tpm2_flush_space(struct tpm_chip *chip)
 	struct tpm_space *space = &chip->work_space;
 	int i;
 
-	if (!space)
-		return;
-
 	for (i = 0; i < ARRAY_SIZE(space->context_tbl); i++)
 		if (space->context_tbl[i] && ~space->context_tbl[i])
 			tpm2_flush_context(chip, space->context_tbl[i]);
Quoted 7 lines
Quoted 4 lines

base-commit: 949692da7211572fac419b2986b6abc0cd1aeb76

Signature · 1 line
2.47.3

It's all good otherwise, just need clarification as we are learning how to deal with these patches :-)

Hold on, I would not remove it: it's an invariant. It's not dead code.

BR, Jarkko

#7 lore raw
In reply to Jarkko Sakkinen
Quoted 3 lines · Jarkko Sakkinen

Just for sake of understanding:

What is "kiro" and is assisted-by the tag supposed to be used here?

Kiro is an AI-powered IDE built on top of Claude [1]. In this case, it helped with reasoning about the correctness of the fix, generating test cases to validate the change, and formalizing the commit message.

I'm trying to follow the established guidelines for AI attribution in kernel contributions to maintain scientific integrity [2] ;-)

On a more utopian note, it is likely that in a year or so there'll be research conducted into how AI-assisted development is influencing the changes to the kernel. My goal is to get into the habit of indicating that AI was used to increase the number of valid data points.

[1] https://kiro.dev/
[2] https://docs.kernel.org/process/coding-assistants.html

Gunnar

maintainer
#8 lore raw
Quoted 11 lines · Gunnar Kudrjavets
Quoted 3 lines · Jarkko Sakkinen

Just for sake of understanding:

What is "kiro" and is assisted-by the tag supposed to be used here?

Kiro is an AI-powered IDE built on top of Claude [1]. In this case, it helped with reasoning about the correctness of the fix, generating test cases to validate the change, and formalizing the commit message.

I'm trying to follow the established guidelines for AI attribution in kernel contributions to maintain scientific integrity [2] ;-)

OK, cool. I do know what Claude is :-) Just was unfamiliar term.

Quoted 5 lines

On a more utopian note, it is likely that in a year or so there'll be research conducted into how AI-assisted development is influencing the changes to the kernel. My goal is to get into the habit of indicating that AI was used to increase the number of valid data points.

Sure.

Quoted 5 lines

[1] https://kiro.dev/
[2] https://docs.kernel.org/process/coding-assistants.html

Gunnar

BR, Jarkko