Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   "Fabio M. De Francesco" <fmdefrancesco@gmail.com>
To:     Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Fabio Aiuto <fabioaiuto83@gmail.com>
Cc:     outreachy-kernel@googlegroups.com, linux-staging@lists.linux.dev,
        linux-kernel@vger.kernel.org
Subject: Re: [Outreachy kernel] [PATCH 1/2] staging: rtl8723bs: hal: Remove set but unused variables
Date:   Thu, 29 Apr 2021 12:38:09 +0200
Message-ID: <1870536.WTzYeaBcXf@linux.local>
In-Reply-To: <20210429102201.GF1409@agape.jhs>
References: <20210428113346.28305-1-fmdefrancesco@gmail.com> <YIqECQk/1Zb4JRjb@kroah.com> <20210429102201.GF1409@agape.jhs>
MIME-Version: 1.0
Content-Transfer-Encoding: 7Bit
Content-Type: text/plain; charset="us-ascii"
Precedence: bulk

On Thursday, April 29, 2021 12:22:02 PM CEST Fabio Aiuto wrote:
> On Thu, Apr 29, 2021 at 12:01:45PM +0200, Greg Kroah-Hartman wrote:
> > On Thu, Apr 29, 2021 at 10:25:53AM +0200, Fabio Aiuto wrote:
> > > On Thu, Apr 29, 2021 at 09:44:47AM +0200, Fabio M. De Francesco wrote:
> > > > On Thursday, April 29, 2021 9:26:20 AM CEST Fabio Aiuto wrote:
> > > > > Hi Fabio,
> > > > > 
> > > > > On Wed, Apr 28, 2021 at 01:33:45PM +0200, Fabio M. De Francesco 
wrote:
> > > > > > Removed four set but unused variables. Issue detected by gcc.
> > > > > > 
> > > > > > Signed-off-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>
> > > > > > ---
> > > > > > 
> > > > > >  drivers/staging/rtl8723bs/hal/rtl8723b_hal_init.c | 5 -----
> > > > > >  1 file changed, 5 deletions(-)
> > > > > > 
> > > > > > diff --git a/drivers/staging/rtl8723bs/hal/rtl8723b_hal_init.c
> > > > > > b/drivers/staging/rtl8723bs/hal/rtl8723b_hal_init.c index
> > > > 
> > > > 082448557b53..96cb4426a0f4
> > > > 
> > > > > > 100644
> > > > > > --- a/drivers/staging/rtl8723bs/hal/rtl8723b_hal_init.c
> > > > > > +++ b/drivers/staging/rtl8723bs/hal/rtl8723b_hal_init.c
> > > > > > @@ -3900,14 +3900,11 @@ u8 GetHalDefVar8723B(struct adapter 
*padapter,
> > > > 
> > > > enum
> > > > 
> > > > > > hal_def_variable variable, v>
> > > > > > 
> > > > > >  			u32 cmd;
> > > > > >  			u32 ra_info1, ra_info2;
> > > > > >  			u32 rate_mask1, rate_mask2;
> > > > > > 
> > > > > > -			u8 curr_tx_rate, curr_tx_sgi, hight_rate,
> > > > 
> > > > lowest_rate;
> > > > 
> > > > > >  			cmd = 0x40000100 | mac_id;
> > > > > >  			rtw_write32(padapter,
> > > > 
> > > > REG_HMEBOX_DBG_2_8723B, cmd);
> > > > 
> > > > > >  			msleep(10);
> > > > > >  			ra_info1 = rtw_read32(padapter, 0x2F0);
> > > > > > 
> > > > > > -			curr_tx_rate = ra_info1&0x7F;
> > > > > > -			curr_tx_sgi = (ra_info1>>7)&0x01;
> > > > > > 
> > > > > >  			cmd = 0x40000400 | mac_id;
> > > > > >  			rtw_write32(padapter,
> > > > 
> > > > REG_HMEBOX_DBG_2_8723B, cmd);
> > > > 
> > > > > > @@ -3916,8 +3913,6 @@ u8 GetHalDefVar8723B(struct adapter 
*padapter, enum
> > > > > > hal_def_variable variable, v>
> > > > > > 
> > > > > >  			ra_info2 = rtw_read32(padapter, 0x2F4);
> > > > > >  			rate_mask1 = rtw_read32(padapter, 0x2F8);
> > > > > >  			rate_mask2 = rtw_read32(padapter, 0x2FC);
> > > > > > 
> > > > > > -			hight_rate = ra_info2&0xFF;
> > > > > > -			lowest_rate = (ra_info2>>8)  & 0xFF;
> > > > > > 
> > > > > >  		}
> > > > > >  		break;
> > > > > 
> > > > > rate_mask{1,2} and ra_info{1,2} seems to be unused as well.
> > > > > 
> > > > > thank you,
> > > > > 
> > > > > fabio
> > > > 
> > > > Hello Fabio,
> > > > 
> > > > I'm not sure about it: rtw_read32 calls a pointer to a function. I'm 
don't
> > > > know drivers programming, however that function looks like an 
implementation
> > > > of a read() system call. So I wouldn't be so sure to remove those 
calls.
> > > > 
> > > > Could calling a (*read) method have side effects on subsequent read()? 
I mean:
> > > > could it update some internal data structure? If not I can remove the
> > > > variables you mentioned above and the calls to read32.
> > > > 
> > > > I'm looking forward to read your reply.
> > > > 
> > > > Thanks,
> > > > 
> > > > Fabio
> > > 
> > > hi Fabio,
> > > 
> > > rtw_read32 is a macro wrapping _rtw_read32() defined as follows (in 
core/rtw_io.c):
> > > 
> > > u32 _rtw_read32(struct adapter *adapter, u32 addr)
> > > {
> > > 
> > >         u32 r_val;
> > >         /* struct       io_queue        *pio_queue = (struct io_queue
> > >         *)adapter->pio_queue; */
> > >         struct io_priv *pio_priv = &adapter->iopriv;
> > >         struct  intf_hdl                *pintfhdl = &(pio_priv->intf);
> > >         u32 (*_read32)(struct intf_hdl *pintfhdl, u32 addr);
> > >         
> > >         _read32 = pintfhdl->io_ops._read32;
> > >         
> > >         r_val = _read32(pintfhdl, addr);
> > >         return rtw_le32_to_cpu(r_val);
> > > 
> > > }
> > > 
> > > the actual read seems to be performed by the handler contained in
> > > 
> > > 	pintfhdl->io_ops._read32;
> > > 
> > > so:
> > > 
> > > $ grep -r '\b_read32' drivers/staging/rtl8723bs/
> > > 
> > > drivers/staging/rtl8723bs/hal/sdio_ops.c:	ops->_read32 = 
&sdio_read32;
> > > 
> > > this is the place where _read32 is stored with sdio_read32 reference...
> > > 
> > > drivers/staging/rtl8723bs/core/rtw_io.c:	u32 (*_read32)(struct 
intf_hdl *pintfhdl,
> > > u32 addr); drivers/staging/rtl8723bs/core/rtw_io.c:	_read32 =
> > > pintfhdl->io_ops._read32;
> > > ...
> > > 
> > > if you check it in hal/sdio_ops.c, nothing is written, just reads are
> > > performed (and it's not odd, for a read function isn't supposed to write
> > > something under the hood ;)).
> > 
> > Yes, but many types of hardware _REQUIRE_ reads to do something.  So
> > "read that does nothing" is a requirement for some operations.
> > 
> > As an example, a write is only guaranteed to have been finished if you
> > do a read of the same location back from it on some hardware busses.
> > The bus can reorder things, but a write/read of the same location can
> > not be reordered.
> > 
> > Sometimes you have to do reads multiple times to get things to "stick".
> > 
> > Other times reading from a location changes a state in the hardware
> > (horrid but HW designers aren't the brightest at times...)
> > 
> > So you can NOT just remove reads without knowing that the hardware does
> > not require this.  This is an issue where GCC "warnings" mean nothing as
> > gcc does not actually know what hardware does, or does not, do for many
> > things.
> > 
> > thanks,
> > 
> > greg k-h
> 
> thank you for explanation, my hardware knowledge is poor:(
> Sorry for noise.
> 
> fabio
>
I suspected that removing those variables could have been a source of troubles 
(but I was thinking of possible side effects on internal kernel's data, not of 
hardware related idiosyncrasies), however I think that you did well to point 
it out because:

1) We learned something new from Greg;
2) I learned that, for the purpose of finding definitions, vim's ctrl-] is not 
the right way to work out the problem.

If you have time, I'd appreciate some comments on the topic of line (2).

Thanks,

Fabio