[PATCH nft] evaluate: init cmd pointer for new on-stack context

[PATCH nft] evaluate: init cmd pointer for new on-stack context

[Florian Westphal]

else, this will segfault when trying to print the
"table 'x' doesn't exist" error message.

Signed-off-by: Florian Westphal <fw@strlen.de>
---
 src/evaluate.c                                   | 1 +
 tests/shell/testcases/chains/0041chain_binding_0 | 6 ++++++
 2 files changed, 7 insertions(+)

diff --git a/src/evaluate.c b/src/evaluate.c
index 2732f5f49e06..07a4b0ad19b0 100644
--- a/src/evaluate.c
+++ b/src/evaluate.c
@@ -3425,6 +3425,7 @@ static int stmt_evaluate_chain(struct eval_ctx *ctx, struct stmt *stmt)
 			struct eval_ctx rule_ctx = {
 				.nft	= ctx->nft,
 				.msgs	= ctx->msgs,
+				.cmd	= ctx->cmd,
 			};
 			struct handle h2 = {};
 
diff --git a/tests/shell/testcases/chains/0041chain_binding_0 b/tests/shell/testcases/chains/0041chain_binding_0
index 59bdbe9f0ba9..806c17535002 100755
--- a/tests/shell/testcases/chains/0041chain_binding_0
+++ b/tests/shell/testcases/chains/0041chain_binding_0
@@ -1,5 +1,11 @@
 #!/bin/bash
 
+# no table z, caused segfault in earlier nft releases
+$NFT insert rule inet x y handle 107 'goto { log prefix "MOO! "; }'
+if [ $? -ne 1 ]; then
+	exit 1
+fi
+
 set -e
 
 EXPECTED="table inet x {
-- 
2.34.1

Re: [PATCH nft] evaluate: init cmd pointer for new on-stack context

[Pablo Neira Ayuso]

On Fri, Mar 04, 2022 at 11:36:34AM +0100, Florian Westphal wrote:
> else, this will segfault when trying to print the
> "table 'x' doesn't exist" error message.

LGTM, thanks. One nitpick below:

> Signed-off-by: Florian Westphal <fw@strlen.de>
> ---
>  src/evaluate.c                                   | 1 +
>  tests/shell/testcases/chains/0041chain_binding_0 | 6 ++++++
>  2 files changed, 7 insertions(+)
> 
> diff --git a/src/evaluate.c b/src/evaluate.c
> index 2732f5f49e06..07a4b0ad19b0 100644
> --- a/src/evaluate.c
> +++ b/src/evaluate.c
> @@ -3425,6 +3425,7 @@ static int stmt_evaluate_chain(struct eval_ctx *ctx, struct stmt *stmt)
>  			struct eval_ctx rule_ctx = {
>  				.nft	= ctx->nft,
>  				.msgs	= ctx->msgs,
> +				.cmd	= ctx->cmd,
>  			};
>  			struct handle h2 = {};
>  
> diff --git a/tests/shell/testcases/chains/0041chain_binding_0 b/tests/shell/testcases/chains/0041chain_binding_0
> index 59bdbe9f0ba9..806c17535002 100755
> --- a/tests/shell/testcases/chains/0041chain_binding_0
> +++ b/tests/shell/testcases/chains/0041chain_binding_0
> @@ -1,5 +1,11 @@
>  #!/bin/bash
>  
> +# no table z, caused segfault in earlier nft releases

this is no table x?

> +$NFT insert rule inet x y handle 107 'goto { log prefix "MOO! "; }'
> +if [ $? -ne 1 ]; then
> +	exit 1
> +fi
> +
>  set -e
>  
>  EXPECTED="table inet x {
> -- 
> 2.34.1
> 

Re: [PATCH nft] evaluate: init cmd pointer for new on-stack context

[Florian Westphal]

Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> On Fri, Mar 04, 2022 at 11:36:34AM +0100, Florian Westphal wrote:
> > else, this will segfault when trying to print the
> > "table 'x' doesn't exist" error message.
> 
> LGTM, thanks. One nitpick below:

I've pushed it out with the corrected comment.

Re: [PATCH nft] evaluate: init cmd pointer for new on-stack context

[Phil Sutter]

On Fri, Mar 04, 2022 at 11:54:39AM +0100, Pablo Neira Ayuso wrote:
> On Fri, Mar 04, 2022 at 11:36:34AM +0100, Florian Westphal wrote:
> > else, this will segfault when trying to print the
> > "table 'x' doesn't exist" error message.
> 
> LGTM, thanks. One nitpick below:

This also fixes the segfault I was trying to avoid with my patch for
string_misspell_update(). I had definitely tried Florian's solution, but
it didn't work for me. Guess there's a bug in my test env or something.
:(

Thanks, Phil