| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20210916125029.GL2116@kadam>
Date: Thu, 16 Sep 2021 15:50:30 +0300
From: Dan Carpenter <dan.carpenter@...cle.com>
To: "Fabio M. De Francesco" <fmdefrancesco@...il.com>
Cc: Larry Finger <Larry.Finger@...inger.net>,
Phillip Potter <phil@...lpotter.co.uk>,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
Pavel Skripkin <paskripkin@...il.com>,
linux-staging@...ts.linux.dev, linux-kernel@...r.kernel.org,
David Laight <david.Laight@...lab.com>
Subject: Re: [PATCH v6 16/19] staging: r8188eu: Clean up rtw_read*() and
rtw_write*()
On Thu, Sep 16, 2021 at 02:14:14PM +0200, Fabio M. De Francesco wrote:
> On Thursday, September 16, 2021 1:36:06 PM CEST Dan Carpenter wrote:
> > On Wed, Sep 15, 2021 at 11:11:00PM +0200, Fabio M. De Francesco wrote:
> > > Clean up rtw_read{8,16,32}() and rtw_write{8,16,32,N}() in
> usb_ops_linux.c.
> > >
> >
> > It would be good to know what you did more specifically.
> >
> > 1) Rename variables:
> > pio_priv => io_priv
> > pintfhdl => intfhdl
> > wvalue => address.
> > 2) Remove unnecessary casts.
> > 3) Fix types. Use __le16 instead of __le32.
>
> Dear Dan,
>
> I'm sorry for missing that. :(
>
> Now I remember that you asked for this specifications at least once (if not
> twice). I'll redo the commit message and add the list above in v7. I guess
> that I have to do the same in 15/19.
>
> > The last one is a small KASan bug fix. So good job on that.
>
> Thanks (even if I don't yet know anything about KASan).
>
> > > Co-developed-by: Pavel Skripkin <paskripkin@...il.com>
> > > Signed-off-by: Pavel Skripkin <paskripkin@...il.com>
> > > Signed-off-by: Fabio M. De Francesco <fmdefrancesco@...il.com>
> > > ---
> > > drivers/staging/r8188eu/hal/usb_ops_linux.c | 68 ++++++++++-----------
> > > 1 file changed, 34 insertions(+), 34 deletions(-)
> > >
> > > diff --git a/drivers/staging/r8188eu/hal/usb_ops_linux.c b/drivers/
> staging/r8188eu/hal/usb_ops_linux.c
> > > index 2098ce935dc0..d87da84eca07 100644
> > > --- a/drivers/staging/r8188eu/hal/usb_ops_linux.c
> > > +++ b/drivers/staging/r8188eu/hal/usb_ops_linux.c
> > > @@ -91,91 +91,91 @@ static int usbctrl_vendorreq(struct intf_hdl
> *intfhdl, u16 value, void *data, u1
> > >
> > > u8 rtw_read8(struct adapter *adapter, u32 addr)
> > > {
> > > - struct io_priv *pio_priv = &adapter->iopriv;
> > > - struct intf_hdl *pintfhdl = &pio_priv->intf;
> > > - u16 wvalue = (u16)(addr & 0x0000ffff);
> > > + struct io_priv *io_priv = &adapter->iopriv;
> > > + struct intf_hdl *intfhdl = &io_priv->intf;
> > > + u16 address = addr & 0xffff;
> > > u8 data;
> > > -
> >
> > Deleting this line introduces a checkpatch warning.
>
> I didn't notice the warning. This too will be fixed in v7.
>
> > > - usbctrl_vendorreq(pintfhdl, wvalue, &data, 1,
> REALTEK_USB_VENQT_READ);
> > > + usbctrl_vendorreq(intfhdl, address, &data, 1,
> REALTEK_USB_VENQT_READ);
> > >
> > > return data;
> > > }
> > >
> > > u16 rtw_read16(struct adapter *adapter, u32 addr)
> > > {
> > > - struct io_priv *pio_priv = &adapter->iopriv;
> > > - struct intf_hdl *pintfhdl = &pio_priv->intf;
> > > - u16 wvalue = (u16)(addr & 0x0000ffff);
> > > - __le32 data;
> > > + struct io_priv *io_priv = &adapter->iopriv;
> > > + struct intf_hdl *intfhdl = &io_priv->intf;
> > > + u16 address = addr & 0xffff;
> > > + __le16 data;
> > >
> > > - usbctrl_vendorreq(pintfhdl, wvalue, &data, 2,
> REALTEK_USB_VENQT_READ);
> > > + usbctrl_vendorreq(intfhdl, address, &data, 2,
> REALTEK_USB_VENQT_READ);
> > >
> > > - return (u16)(le32_to_cpu(data) & 0xffff);
> > > + return le16_to_cpu(data);
> >
> > The last two bytes of "data" are not initialized. I do not think that
> > will cause a bug on either endian type of system during runtime but I
> > this that KASan will catch it and complain.
>
> I don't want to add mistakes on mistakes. I guess that you are talking of the
> same fix you wrote above and that "return le16_to_cpu(data);" is correct.
> Am I interpreting your words in the correct way?
>
In the original code the last two bytes of "data" were uninitialized.
KASan will spot this as a bug, but it doesn't affect runtime because we
mask away those bytes anyway.
> > > }
> > >
> > > u32 rtw_read32(struct adapter *adapter, u32 addr)
> > > {
> > > - struct io_priv *pio_priv = &adapter->iopriv;
> > > - struct intf_hdl *pintfhdl = &pio_priv->intf;
> > > - u16 wvalue = (u16)(addr & 0x0000ffff);
> > > + struct io_priv *io_priv = &adapter->iopriv;
> > > + struct intf_hdl *intfhdl = &io_priv->intf;
> > > + u16 address = addr & 0xffff;
> > > __le32 data;
> > >
> > > - usbctrl_vendorreq(pintfhdl, wvalue, &data, 4,
> REALTEK_USB_VENQT_READ);
> > > + usbctrl_vendorreq(intfhdl, address, &data, 4,
> REALTEK_USB_VENQT_READ);
> > >
> > > return le32_to_cpu(data);
> > > }
> > >
> > > int rtw_write8(struct adapter *adapter, u32 addr, u8 val)
> > > {
> > > - struct io_priv *pio_priv = &adapter->iopriv;
> > > - struct intf_hdl *pintfhdl = &pio_priv->intf;
> > > - u16 wvalue = (u16)(addr & 0x0000ffff);
> > > + struct io_priv *io_priv = &adapter->iopriv;
> > > + struct intf_hdl *intfhdl = &io_priv->intf;
> > > + u16 address = addr & 0xffff;
> > > int ret;
> > >
> > > - ret = usbctrl_vendorreq(pintfhdl, wvalue, &val, 1,
> REALTEK_USB_VENQT_WRITE);
> > > + ret = usbctrl_vendorreq(intfhdl, address, &val, 1,
> REALTEK_USB_VENQT_WRITE);
> > >
> > > return RTW_STATUS_CODE(ret);
> > > }
> > >
> > > int rtw_write16(struct adapter *adapter, u32 addr, u16 val)
> > > {
> > > - struct io_priv *pio_priv = &adapter->iopriv;
> > > - struct intf_hdl *pintfhdl = &pio_priv->intf;
> > > - u16 wvalue = (u16)(addr & 0x0000ffff);
> > > - __le32 data = cpu_to_le32(val & 0x0000ffff);
> > > + struct io_priv *io_priv = &adapter->iopriv;
> > > + struct intf_hdl *intfhdl = &io_priv->intf;
> > > + __le16 data = cpu_to_le16(val);
> >
> > This is the other interesting change. I think the original code works
> > though.
>
> Here too, I'm a bit confused... Do yo prefer the original code or you're
> saying that, although the original code works fine, I made the correct choice
> in changing it? Can you please confirm?
>
Yeah. The original code was buggy but it still worked fine. Ideally
this kind of logic fix would be in a separate patch from the other
"rename a variable" changes.
regards,
dan carpenter
Powered by blists - more mailing lists