Skip to contentSkip to conversation

Signed-off-by: giladreti <[email protected]>

PATCH In review giladreti 4 participants 5 messages Jan 11, 2021

Subsystems (4)
lore
Share & tools
lore

giladreti

author
#0 lore raw

PATCH 1/1 Signed-off-by: giladreti <[email protected]>

Added support for pointer to mem register spilling, to allow the verifier to track pointer to valid memory addresses. Such pointers are returned for example by a successful call of the bpf_ringbuf_reserve helper.

This patch was suggested as a solution by Yonghong Song.

---
 kernel/bpf/verifier.c | 2 ++
 1 file changed, 2 insertions(+)
kernel/bpf/verifier.c +2 -0
index 17270b8404f1..36af69fac591 100644
@@ -2217,6 +2217,8 @@ static bool is_spillable_regtype(enum bpf_reg_type type)
 	case PTR_TO_RDWR_BUF:
 	case PTR_TO_RDWR_BUF_OR_NULL:
 	case PTR_TO_PERCPU_BTF_ID:
+	case PTR_TO_MEM:
+	case PTR_TO_MEM_OR_NULL:
 		return true;
 	default:
 		return false;
Signature · 1 line
2.27.0
#1 lore raw

The Signed-off-by line should be last in the commit message, not first. First line (which becomes e-mail subject) should describe what the commit does (in a short one liner) and where it does it.

So for your patch it could be something like
bpf: support pointer to mem register spilling in verifier

The commit message should be written in present simple tense, not past simple, ie. not "Added support" but "Add support".

Also we need your name and surname in From: header and Signed-off-by:
tag. So instead of
giladreti <[email protected]>
it should be
Gilad Reti <[email protected]>
if that is your real time. If it is not, please provide your real name.

Quoted 22 lines · On Mon, 11 Jan 2021 17:31:23 +0200 giladreti <[email protected]> wrote:

Added support for pointer to mem register spilling, to allow the verifier to track pointer to valid memory addresses. Such pointers are returned for example by a successful call of the bpf_ringbuf_reserve helper.

This patch was suggested as a solution by Yonghong Song.

---
 kernel/bpf/verifier.c | 2 ++
 1 file changed, 2 insertions(+)
kernel/bpf/verifier.c +2 -0
index 17270b8404f1..36af69fac591 100644
@@ -2217,6 +2217,8 @@ static bool is_spillable_regtype(enum bpf_reg_type type)
 	case PTR_TO_RDWR_BUF:
 	case PTR_TO_RDWR_BUF_OR_NULL:
 	case PTR_TO_PERCPU_BTF_ID:
+	case PTR_TO_MEM:
+	case PTR_TO_MEM_OR_NULL:
 		return true;
 	default:
 		return false;
maintainer
#2 lore raw
In reply to giladreti

Hello Gilad,

Quoted 5 lines · giladreti

Added support for pointer to mem register spilling, to allow the verifier to track pointer to valid memory addresses. Such pointers are returned for example by a successful call of the bpf_ringbuf_reserve helper.

This patch was suggested as a solution by Yonghong Song.

The SoB should not be in subject line but as part of the commit message instead and with proper name, e.g.

For subject line, please use a short summary that fits the patch prefixed with the subsystem "bpf: [...]", see also [0] as an example. Thanks.

It would be good if you could also add a BPF selftest for this [1].

[0] https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/commit/?id=e22d7f05e445165e58feddb4e40cc9c0f94453bc
[1] https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/tree/tools/testing/selftests/bpf/
https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/tree/tools/testing/selftests/bpf/verifier/spill_fill.c

Quoted 18 lines
---
  kernel/bpf/verifier.c | 2 ++
  1 file changed, 2 insertions(+)
kernel/bpf/verifier.c +2 -0
index 17270b8404f1..36af69fac591 100644
@@ -2217,6 +2217,8 @@ static bool is_spillable_regtype(enum bpf_reg_type type)
  	case PTR_TO_RDWR_BUF:
  	case PTR_TO_RDWR_BUF_OR_NULL:
  	case PTR_TO_PERCPU_BTF_ID:
+	case PTR_TO_MEM:
+	case PTR_TO_MEM_OR_NULL:
  		return true;
  	default:
  		return false;
author
#3 lore raw
Quoted 15 lines · On Mon, Jan 11, 2021, 17:55 Daniel Borkmann <[email protected]> wrote:

Hello Gilad,

Quoted 5 lines · giladreti

Added support for pointer to mem register spilling, to allow the verifier to track pointer to valid memory addresses. Such pointers are returned for example by a successful call of the bpf_ringbuf_reserve helper.

This patch was suggested as a solution by Yonghong Song.

The SoB should not be in subject line but as part of the commit message instead and with proper name, e.g.

Quoted 9 lines

For subject line, please use a short summary that fits the patch prefixed with the subsystem "bpf: [...]", see also [0] as an example. Thanks.

It would be good if you could also add a BPF selftest for this [1].

   [0] https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/commit/?id=e22d7f05e445165e58feddb4e40cc9c0f94453bc
   [1] https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/tree/tools/testing/selftests/bpf/
       https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/tree/tools/testing/selftests/bpf/verifier/spill_fill.c

Sure. Thanks for your guidance. As you can probably tell, I am new to kernel code contribution (in fact this is a first time for me). Should I try to submit this patch again?

Sorry in advance for all the overhead I may be causing to you...

Quoted 19 lines
Quoted 18 lines
---
  kernel/bpf/verifier.c | 2 ++
  1 file changed, 2 insertions(+)
kernel/bpf/verifier.c +2 -0
index 17270b8404f1..36af69fac591 100644
@@ -2217,6 +2217,8 @@ static bool is_spillable_regtype(enum bpf_reg_type type)
      case PTR_TO_RDWR_BUF:
      case PTR_TO_RDWR_BUF_OR_NULL:
      case PTR_TO_PERCPU_BTF_ID:
+     case PTR_TO_MEM:
+     case PTR_TO_MEM_OR_NULL:
              return true;
      default:
              return false;
#4 lore raw
Quoted 22 lines · Gilad Reti
Quoted 20 lines · On Mon, Jan 11, 2021, 17:55 Daniel Borkmann <[email protected]> wrote:

Hello Gilad,

Quoted 5 lines · giladreti

Added support for pointer to mem register spilling, to allow the verifier to track pointer to valid memory addresses. Such pointers are returned for example by a successful call of the bpf_ringbuf_reserve helper.

This patch was suggested as a solution by Yonghong Song.

The SoB should not be in subject line but as part of the commit message instead and with proper name, e.g.

For subject line, please use a short summary that fits the patch prefixed with the subsystem "bpf: [...]", see also [0] as an example. Thanks.

It would be good if you could also add a BPF selftest for this [1].

Quoted 8 lines
Quoted 4 lines
   [0] https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/commit/?id=e22d7f05e445165e58feddb4e40cc9c0f94453bc
   [1] https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/tree/tools/testing/selftests/bpf/
       https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/tree/tools/testing/selftests/bpf/verifier/spill_fill.c

Sure. Thanks for your guidance. As you can probably tell, I am new to kernel code contribution (in fact this is a first time for me). Should I try to submit this patch again?

In addition to all already mentioned things, also make sure you have [PATCH bpf] prefix in the subject, to identify that this is a bug fix for the bpf tree.

Also you missed adding Fixes tag, please add this:

Fixes 457f44363a88 ("bpf: Implement BPF ring buffer and verifier

support for it")

And yes, please re-submit with all the feedback incorporated (including the selftest).

Quoted 22 lines

Sorry in advance for all the overhead I may be causing to you...

Quoted 19 lines
Quoted 18 lines
---
  kernel/bpf/verifier.c | 2 ++
  1 file changed, 2 insertions(+)
kernel/bpf/verifier.c +2 -0
index 17270b8404f1..36af69fac591 100644
@@ -2217,6 +2217,8 @@ static bool is_spillable_regtype(enum bpf_reg_type type)
      case PTR_TO_RDWR_BUF:
      case PTR_TO_RDWR_BUF_OR_NULL:
      case PTR_TO_PERCPU_BTF_ID:
+     case PTR_TO_MEM:
+     case PTR_TO_MEM_OR_NULL:
              return true;
      default:
              return false;