Re: [nft PATCH v4 05/32] ct: support `NULL` symbol-tables when looking up labels

[Jeremy Sowden <jeremy@azazel.net>]

[-- Attachment #1: Type: text/plain, Size: 1563 bytes --]

On 2022-04-05, at 13:15:45 +0200, Florian Westphal wrote:
> Jeremy Sowden <jeremy@azazel.net> wrote:
> > If the symbol-table passed to `ct_label2str` is `NULL`, return `NULL`.
> 
> It would be nice to describe why, not what.
> Does this fix a crash when the label file is missing?

I've been putting debug logging in various places while developing and
testing things.  Here's a function for dumping expressions to stderr:

  static inline void
  dbg_print_expr(const char *func, const char *name,
                 const struct expr *expr)
  {
    struct output_ctx *dbgctx = &(struct output_ctx) { .output_fp = stderr };

    nft_print(dbgctx, "%s: %s = (%s, %u, %s, %s) { ",
              func, name,
              expr_ops(expr)->name,
              expr->len,
              expr->dtype ? expr->dtype->name : "",
              byteorder_names[expr->byteorder]);
    expr_print(expr, dbgctx);
    nft_print(dbgctx, " }\n");
  }

There are two problems with this.  Firstly, the `byteorder_names` array
is defined in src/evaluate.c and so this function cannot be used
elsewhere.  Secondly, the symbol-tables in the output context are not
initialized, which results in crashes when trying to print symbolic
constants and CT labels.  Patch 3 fixes the former problem by moving the
`byteorder_names` array, and patches 4 & 5 fix the latter problem by
adding NULL checks before trying to dereference the symbol-tables.  They
seemed like small, low-impact changes that could be upstreamed, so that
I didn't have to carry them.

J.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]