On Mon, Jul 31, 2023 at 04:57:18PM +0800, Peng Zhang wrote:
>
>
> 在 2023/7/31 14:34, Dan Carpenter 写道:
> > Hello Peng Zhang,
> >
> > This is a semi-automatic email about new static checker warnings.
> >
> > The patch d126b5d9410f: "maple_tree: make mas_validate_gaps() to
> > check metadata" from Jul 11, 2023, leads to the following Smatch
> > complaint:
> >
> > tools/testing/radix-tree/../../../lib/maple_tree.c:6989 mas_validate_gaps()
> > warn: variable dereferenced before check 'gaps' (see line 6983)
> >
> > tools/testing/radix-tree/../../../lib/maple_tree.c
> > 6982
> > 6983 if (gaps[offset] != max_gap) {
> > ^^^^^
> > Dereferenced.
> >
> > 6984 pr_err("gap %p[%u] is not the largest gap %lu\n",
> > 6985 node, offset, max_gap);
> > 6986 MT_BUG_ON(mas->tree, 1);
> > 6987 }
> > 6988
> > 6989 MT_BUG_ON(mas->tree, !gaps);
> > ^^^^^
> > Checked too late. This is pointless as well. Just delete this line.
> Since this is a validator, it is only used for testing, so it will not
> have any impact on normal functionality. I'm not going to delete this
> line, will move it before dereferencing. Thank you for your report.
It will probably still end up generating a warning because MT_BUG_ON()
doesn't actually call BUG(). Smatch actually generates two warnings for
this file.
tools/testing/radix-tree/../../../lib/maple_tree.c:6983 mas_validate_gaps() error: we previously assumed 'gaps' could be null (see line 6950)
tools/testing/radix-tree/../../../lib/maple_tree.c:6989 mas_validate_gaps() warn: variable dereferenced before check 'gaps' (see line 6983)
That business of writing crappier code for the testing/ directory is a
weird thing because it has to be done so specifically to not generate
warnings and debate. It's sometimes easier in the end to just do it
correctly.
regards,
dan carpenter
