Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <202209170232.grVMomkJ-lkp@intel.com>
In-Reply-To: <202209170232.grVMomkJ-lkp@intel.com>
From:   Linus Walleij <linus.walleij@linaro.org>
Date:   Sun, 18 Sep 2022 16:33:33 +0200
Message-ID: <CACRpkdYGMToSn1XR0Zrvv3yKeOX7_tw0=y0NKhOXN7O8KL-Ntg@mail.gmail.com>
Subject: Re: [linux-stable-rc:linux-5.4.y 5364/5387] drivers/infiniband/sw/siw/siw_qp_tx.c:32:37:
 warning: cast to pointer from integer of different size
To:     kernel test robot <lkp@intel.com>
Cc:     kbuild-all@lists.01.org, linux-kernel@vger.kernel.org,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Leon Romanovsky <leon@kernel.org>,
        Sasha Levin <sashal@kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, Sep 16, 2022 at 8:19 PM kernel test robot <lkp@intel.com> wrote:

> commit: e7f78835d551bb2deb5aa3346d84c8f03ade313d [5364/5387] RDMA/siw: Pass a pointer to virt_to_page()
> config: riscv-randconfig-r013-20220916 (https://download.01.org/0day-ci/archive/20220917/202209170232.grVMomkJ-lkp@intel.com/config)
> compiler: riscv32-linux-gcc (GCC) 12.1.0

So if RISCV32 ...

>     31          if (paddr)
>   > 32                  return virt_to_page((void *)paddr);

Think that passing a (void *) to virt_to_page() is a problem:

> warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
>       32 |                 return virt_to_page((void *)paddr);

I would think this is a problem with RISCV32.

What has this arch done?

This is a few lines up:

> dma_addr_t paddr = siw_pbl_get_buffer(pbl, offset, NULL, idx);

dma_addr_t is a different size than (void *)?

Given that the patch raising this problem looked like this:

-               return virt_to_page(paddr);
+               return virt_to_page((void *)paddr);

I doubt that it has created a bug that wasn't there before. Passing
a dma_addr_t to virt_to_page() might be wrong, but that is what the
driver has been doing all the time.

siw_pbl_get_buffer() claims to
"Gets physical address backed by PBL element."
Why is that even passed to virt_to_page() if it's no virtual address
but a dma_addr_t to begin with?

I certainly don't understand SIW, but this bug, if it is a bug (or a type and
documentation bug) is clearly not new.

Yours,
Linus Walleij