Re: [PATCH] RDMA/cxgb4: Replace 0-length arrays with flexible arrays

[Leon Romanovsky <leon@kernel.org>]

On Fri, Jan 13, 2023 at 05:03:48PM -0800, Kees Cook wrote:
> On Sun, Jan 08, 2023 at 02:08:52PM +0200, Leon Romanovsky wrote:
> > On Thu, Jan 05, 2023 at 02:32:32PM -0800, Kees Cook wrote:
> > > Zero-length arrays are deprecated[1]. Replace all remaining
> > > 0-length arrays with flexible arrays. Detected with GCC 13, using
> > > -fstrict-flex-arrays=3:
> > > 
> > > In function 'build_rdma_write',
> > >     inlined from 'c4iw_post_send' at ../drivers/infiniband/hw/cxgb4/qp.c:1173:10:
> > > ../drivers/infiniband/hw/cxgb4/qp.c:597:38: warning: array subscript 0 is outside array bounds of 'struct fw_ri_immd[0]' [-Warray-bounds=]
> > >   597 |                 wqe->write.u.immd_src[0].r2 = 0;
> > >       |                 ~~~~~~~~~~~~~~~~~~~~~^~~
> > > ../drivers/infiniband/hw/cxgb4/t4fw_ri_api.h: In function 'c4iw_post_send':
> > > ../drivers/infiniband/hw/cxgb4/t4fw_ri_api.h:567:35: note: while referencing 'immd_src'
> > >   567 |                 struct fw_ri_immd immd_src[0];
> > >       |                                   ^~~~~~~~
> > > 
> > > Additionally drop the unused C99_NOT_SUPPORTED ifndef lines.
> > > 
> > > [1] https://www.kernel.org/doc/html/latest/process/deprecated.html#zero-length-and-one-element-arrays
> > > 
> > > Cc: Potnuri Bharat Teja <bharat@chelsio.com>
> > > Cc: Jason Gunthorpe <jgg@ziepe.ca>
> > > Cc: Leon Romanovsky <leon@kernel.org>
> > > Cc: "Gustavo A. R. Silva" <gustavoars@kernel.org>
> > > Cc: linux-rdma@vger.kernel.org
> > > Signed-off-by: Kees Cook <keescook@chromium.org>
> > > ---
> > >  drivers/infiniband/hw/cxgb4/t4fw_ri_api.h | 26 ++++++-----------------
> > >  1 file changed, 6 insertions(+), 20 deletions(-)
> > 
> > <...>
> > 
> > >  #define FW_RI_SEND_WR_SENDOP_S		0
> > > @@ -618,12 +606,10 @@ struct fw_ri_rdma_write_cmpl_wr {
> > >  		struct fw_ri_isgl isgl_src;
> > >  	} u_cmpl;
> > >  	__be64 r3;
> > > -#ifndef C99_NOT_SUPPORTED
> > >  	union fw_ri_write {
> > > -		struct fw_ri_immd immd_src[0];
> > > -		struct fw_ri_isgl isgl_src[0];
> > > +		DECLARE_FLEX_ARRAY(struct fw_ri_immd, immd_src);
> > > +		DECLARE_FLEX_ARRAY(struct fw_ri_isgl, isgl_src);
> > 
> > smatch built from commit 40351132df3b ("strlen: add __builtin and
> > __fortify functions") produces the following warning:
> > drivers/infiniband/hw/cxgb4/t4fw_ri_api.h:575:17: warning: array of flexible structures
> > 
> > Is it expected? What will prevent from getting this warning from 0-day
> > build bots?
> 
> Err, I'm not sure I understand? Does smatch define "C99_NOT_SUPPORTED"?

Why should it define? You removed C99_NOT_SUPPORTED.

Just download smatch, build from scratch and run over this patch. It
generates new warning.

➜  kernel git:(wip/leon-for-next) mkt ci --no-sparse --no-extra-warnings
ccdbefcf661e (HEAD -> build) RDMA/cxgb4: Replace 0-length arrays with flexible arrays
WARNING: Possible unwrapped commit description (prefer a maximum 75 chars per line)
#11:
    inlined from 'c4iw_post_send' at ../drivers/infiniband/hw/cxgb4/qp.c:1173:10:

drivers/infiniband/hw/cxgb4/t4fw_ri_api.h:558:17: warning: array of flexible structures
....



> 
> -- 
> Kees Cook