Skip to contentSkip to conversation

bpf: Don't use %pK through printk

PATCH Applied Thomas Weißschuh 3 participants 5 messages Aug 11, 2025 – Aug 13, 2025

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

PATCH 1/1 bpf: Don't use %pK through printk

In the past %pK was preferable to %p as it would not leak raw pointer values into the kernel log.
Since commit ad67b74d2469 ("printk: hash addresses printed with %p") the regular %p has been improved to avoid this issue. Furthermore, restricted pointers ("%pK") were never meant to be used through printk(). They can still unintentionally leak raw pointers or acquire sleeping locks in atomic contexts.

Switch to the regular pointer formatting which is safer and easier to reason about.

---
 include/linux/filter.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
include/linux/filter.h +1 -1
index 1e7fd3ee759e07534eee7d8b48cffa1dfea056fb..52fecb7a1fe36d233328aabbe5eadcbd7e07cc5a 100644
@@ -1296,7 +1296,7 @@ void bpf_jit_prog_release_other(struct bpf_prog *fp, struct bpf_prog *fp_other);
 static inline void bpf_jit_dump(unsigned int flen, unsigned int proglen,
 				u32 pass, void *image)
 {
-	pr_err("flen=%u proglen=%u pass=%u image=%pK from=%s pid=%d\n", flen,
+	pr_err("flen=%u proglen=%u pass=%u image=%p from=%s pid=%d\n", flen,
 	       proglen, pass, image, current->comm, task_pid_nr(current));
 
 	if (image)

---
base-commit: 8f5ae30d69d7543eee0d70083daf4de8fe15d585
change-id: 20250811-restricted-pointers-bpf-04da04ea1b8a

Best regards,

Signature · 1 line
Thomas Weißschuh <[email protected]>
#1 lore raw
Quoted 22 lines · Thomas Weißschuh

In the past %pK was preferable to %p as it would not leak raw pointer values into the kernel log.
Since commit ad67b74d2469 ("printk: hash addresses printed with %p") the regular %p has been improved to avoid this issue. Furthermore, restricted pointers ("%pK") were never meant to be used through printk(). They can still unintentionally leak raw pointers or acquire sleeping locks in atomic contexts.

Switch to the regular pointer formatting which is safer and easier to reason about.

---
 include/linux/filter.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
include/linux/filter.h +0 -0
index 1e7fd3ee759e07534eee7d8b48cffa1dfea056fb..52fecb7a1fe36d233328aabbe5eadcbd7e07cc5a 100644

include/linux/filter.h void bpf_jit_prog_release_other(struct bpf_prog *fp, struct bpf_prog *fp_other); · line 1300

Quoted 6 lines
include/linux/filter.h +1 -1
@@ -1296,7 +1296,7 @@ void bpf_jit_prog_release_other(struct bpf_prog *fp, struct bpf_prog *fp_other);
 static inline void bpf_jit_dump(unsigned int flen, unsigned int proglen,
                                u32 pass, void *image)
 {
-       pr_err("flen=%u proglen=%u pass=%u image=%pK from=%s pid=%d\n", flen,
+       pr_err("flen=%u proglen=%u pass=%u image=%p from=%s pid=%d\n", flen,
               proglen, pass, image, current->comm, task_pid_nr(current));

this particular printk won't ever be called from atomic context, so I don't think the leak from atomic context matters much here. On the other hand, %pK behavior is controlled by kptr_restrict and that might be useful for debugging, so not sure there is much of a benefit to switching to always obfuscated %p? Or am I missing something else here?

Quoted 11 lines
include/linux/filter.h +0 -1
        if (image)
 
---

base-commit: 8f5ae30d69d7543eee0d70083daf4de8fe15d585
change-id: 20250811-restricted-pointers-bpf-04da04ea1b8a

Best regards,
--
Thomas Weißschuh <[email protected]>

#2 lore raw
Quoted 29 lines · Andrii Nakryiko
Quoted 27 lines · Thomas Weißschuh

In the past %pK was preferable to %p as it would not leak raw pointer values into the kernel log.
Since commit ad67b74d2469 ("printk: hash addresses printed with %p") the regular %p has been improved to avoid this issue. Furthermore, restricted pointers ("%pK") were never meant to be used through printk(). They can still unintentionally leak raw pointers or acquire sleeping locks in atomic contexts.

Switch to the regular pointer formatting which is safer and easier to reason about.

---
 include/linux/filter.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
include/linux/filter.h +1 -1
index 1e7fd3ee759e07534eee7d8b48cffa1dfea056fb..52fecb7a1fe36d233328aabbe5eadcbd7e07cc5a 100644
@@ -1296,7 +1296,7 @@ void bpf_jit_prog_release_other(struct bpf_prog *fp, struct bpf_prog *fp_other);
 static inline void bpf_jit_dump(unsigned int flen, unsigned int proglen,
                                u32 pass, void *image)
 {
-       pr_err("flen=%u proglen=%u pass=%u image=%pK from=%s pid=%d\n", flen,
+       pr_err("flen=%u proglen=%u pass=%u image=%p from=%s pid=%d\n", flen,
Quoted 8 lines

include/linux/filter.h void bpf_jit_prog_release_other(struct bpf_prog *fp, struct bpf_prog *fp_other); · line 1300

Quoted 1 line
include/linux/filter.h +0 -0
               proglen, pass, image, current->comm, task_pid_nr(current));

this particular printk won't ever be called from atomic context, so I don't think the leak from atomic context matters much here. On the other hand, %pK behavior is controlled by kptr_restrict and that might be useful for debugging, so not sure there is much of a benefit to switching to always obfuscated %p? Or am I missing something else here?

As %pK is so easy to abuse and the breakage is very non-obvious, I want to rework it to enforce its usage from "file context".
For that, the printk users need to be gone first.
For debugging, there is still "no_hash_pointers".

How would the image pointer be used for debugging? It is logged from nowhere else.
And the raw image is dumped right after anyways.

Quoted 12 lines
Quoted 12 lines
include/linux/filter.h +0 -1
        if (image)
 
---

base-commit: 8f5ae30d69d7543eee0d70083daf4de8fe15d585
change-id: 20250811-restricted-pointers-bpf-04da04ea1b8a

Best regards,
--
Thomas Weißschuh <[email protected]>

#3 lore raw
Quoted 40 lines · Thomas Weißschuh
Quoted 29 lines · Andrii Nakryiko
Quoted 27 lines · Thomas Weißschuh

In the past %pK was preferable to %p as it would not leak raw pointer values into the kernel log.
Since commit ad67b74d2469 ("printk: hash addresses printed with %p") the regular %p has been improved to avoid this issue. Furthermore, restricted pointers ("%pK") were never meant to be used through printk(). They can still unintentionally leak raw pointers or acquire sleeping locks in atomic contexts.

Switch to the regular pointer formatting which is safer and easier to reason about.

---
 include/linux/filter.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
include/linux/filter.h +1 -1
index 1e7fd3ee759e07534eee7d8b48cffa1dfea056fb..52fecb7a1fe36d233328aabbe5eadcbd7e07cc5a 100644
@@ -1296,7 +1296,7 @@ void bpf_jit_prog_release_other(struct bpf_prog *fp, struct bpf_prog *fp_other);
 static inline void bpf_jit_dump(unsigned int flen, unsigned int proglen,
                                u32 pass, void *image)
 {
-       pr_err("flen=%u proglen=%u pass=%u image=%pK from=%s pid=%d\n", flen,
+       pr_err("flen=%u proglen=%u pass=%u image=%p from=%s pid=%d\n", flen,
Quoted 8 lines

include/linux/filter.h void bpf_jit_prog_release_other(struct bpf_prog *fp, struct bpf_prog *fp_other); · line 1300

Quoted 1 line
include/linux/filter.h +0 -0
               proglen, pass, image, current->comm, task_pid_nr(current));

this particular printk won't ever be called from atomic context, so I don't think the leak from atomic context matters much here. On the other hand, %pK behavior is controlled by kptr_restrict and that might be useful for debugging, so not sure there is much of a benefit to switching to always obfuscated %p? Or am I missing something else here?

Quoted 6 lines

As %pK is so easy to abuse and the breakage is very non-obvious, I want to rework it to enforce its usage from "file context".
For that, the printk users need to be gone first.
For debugging, there is still "no_hash_pointers".

How would the image pointer be used for debugging?

I chatted with Daniel about this offline, and it seems like this is not that critical nowadays. So I went ahead and applied the patch to bpf-next.

Quoted 15 lines

It is logged from nowhere else.
And the raw image is dumped right after anyways.

Quoted 12 lines
Quoted 12 lines
include/linux/filter.h +0 -1
        if (image)
 
---

base-commit: 8f5ae30d69d7543eee0d70083daf4de8fe15d585
change-id: 20250811-restricted-pointers-bpf-04da04ea1b8a

Best regards,
--
Thomas Weißschuh <[email protected]>

#4 lore raw
In reply to Thomas Weißschuh
This patch was applied to bpf/bpf-next.git (master) by Andrii Nakryiko <[email protected]>:

Hello:

This patch was applied to bpf/bpf-next.git (master) by Andrii Nakryiko <[email protected]>:

Quoted 9 lines · On Mon, 11 Aug 2025 14:08:04 +0200 you wrote:

In the past %pK was preferable to %p as it would not leak raw pointer values into the kernel log.
Since commit ad67b74d2469 ("printk: hash addresses printed with %p") the regular %p has been improved to avoid this issue. Furthermore, restricted pointers ("%pK") were never meant to be used through printk(). They can still unintentionally leak raw pointers or acquire sleeping locks in atomic contexts.

[...]

Here is the summary with links:
- bpf: Don't use %pK through printk

    https://git.kernel.org/bpf/bpf-next/c/2caa6b88e0ba

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html