linux-kernel - Re: [PATCH RFC 1/3] staging: r8188eu: add proper rtw

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

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Sat, 21 Aug 2021 07:55:38 +0200
From:   "Fabio M. De Francesco" <fmdefrancesco@...il.com>
To:     Larry.Finger@...inger.net, phil@...lpotter.co.uk,
        gregkh@...uxfoundation.org, straube.linux@...il.com,
        Pavel Skripkin <paskripkin@...il.com>
Cc:     linux-staging@...ts.linux.dev, linux-kernel@...r.kernel.org,
        Pavel Skripkin <paskripkin@...il.com>
Subject: Re: [PATCH RFC 1/3] staging: r8188eu: add proper rtw_read* error handling

On Friday, August 20, 2021 7:07:36 PM CEST Pavel Skripkin wrote:
> rtw_read*() functions call usb_read* inside. These functions could fail
> in some cases; for example: failed to receive control message. These
> cases should be handled to prevent uninit value bugs, since usb_read*
> functions blindly return stack variable without checking if this value
> _actualy_ initialized.
> 
> To achive it, all usb_read* and rtw_read*() argument list is expanded

Dear Pavel,

Please, achive --> achieve.

> with pointer to error and added error usbctrl_vendorreq() error checking.
> If transfer is successful error will be initialized to 0 otherwise to
> error returned from usb_control_msg().
> 
> To not break the build, added error checking for rtw_read*() call all
> across the driver.
> 
> Signed-off-by: Pavel Skripkin <paskripkin@...il.com>
> ---
>  drivers/staging/r8188eu/core/rtw_debug.c      |  79 +++-
>  drivers/staging/r8188eu/core/rtw_efuse.c      |  83 +++-
>  drivers/staging/r8188eu/core/rtw_io.c         |  18 +-
>  drivers/staging/r8188eu/core/rtw_mp.c         |  37 +-
>  drivers/staging/r8188eu/core/rtw_mp_ioctl.c   |  20 +-
>  drivers/staging/r8188eu/core/rtw_pwrctrl.c    |   6 +-
>  drivers/staging/r8188eu/core/rtw_sreset.c     |   7 +-
>  drivers/staging/r8188eu/hal/HalPwrSeqCmd.c    |   9 +-
>  drivers/staging/r8188eu/hal/hal_com.c         |  22 +-
>  drivers/staging/r8188eu/hal/odm_interface.c   |  12 +-
>  drivers/staging/r8188eu/hal/rtl8188e_cmd.c    |  37 +-
>  drivers/staging/r8188eu/hal/rtl8188e_dm.c     |   6 +-
>  .../staging/r8188eu/hal/rtl8188e_hal_init.c   | 198 +++++++--
>  drivers/staging/r8188eu/hal/rtl8188e_phycfg.c |  26 +-
>  drivers/staging/r8188eu/hal/rtl8188e_sreset.c |  20 +-
>  drivers/staging/r8188eu/hal/rtl8188eu_led.c   |  17 +-
>  drivers/staging/r8188eu/hal/usb_halinit.c     | 394 ++++++++++++++----
>  drivers/staging/r8188eu/hal/usb_ops_linux.c   |  16 +-
>  drivers/staging/r8188eu/include/rtw_io.h      |  18 +-
>  drivers/staging/r8188eu/os_dep/ioctl_linux.c  | 168 +++++---
>  20 files changed, 941 insertions(+), 252 deletions(-)

I agree with Philip: please, split this long patch. If I were you, I'd make 
one patch for each of the three rtw_read*() and a fourth patch for usb_read*().

> --- a/drivers/staging/r8188eu/core/rtw_io.c
> +++ b/drivers/staging/r8188eu/core/rtw_io.c
> @@ -34,44 +34,44 @@ jackson@...ltek.com.tw
>  #define rtw_cpu_to_le16(val)		cpu_to_le16(val)
>  #define rtw_cpu_to_le32(val)		cpu_to_le32(val)

Not related to your patch, these macros are useless and misleading.

> -u8 _rtw_read8(struct adapter *adapter, u32 addr)
> +u8 _rtw_read8(struct adapter *adapter, u32 addr, int *error)
>  {
>  	u8 r_val;
>  	struct io_priv *pio_priv = &adapter->iopriv;
>  	struct	intf_hdl *pintfhdl = &pio_priv->intf;
> -	u8 (*_read8)(struct intf_hdl *pintfhdl, u32 addr);
> +	u8 (*_read8)(struct intf_hdl *pintfhdl, u32 addr, int *error);
>  
>  
>  	_read8 = pintfhdl->io_ops._read8;
> -	r_val = _read8(pintfhdl, addr);
> +	r_val = _read8(pintfhdl, addr, error);
>  
>  	return r_val;
>  }

I really don't like passing errors through arguments. Why don't you pass 
a storage location where the function save the byte read and instead use the 
return for errors? I think that this would result in a cleaner design. Furthermore,
it is used everywhere in the kernel.
  
> -u16 _rtw_read16(struct adapter *adapter, u32 addr)
> +u16 _rtw_read16(struct adapter *adapter, u32 addr, int *error)
>  {
>  	u16 r_val;
>  	struct io_priv *pio_priv = &adapter->iopriv;
>  	struct	intf_hdl		*pintfhdl = &pio_priv->intf;
> -	u16 (*_read16)(struct intf_hdl *pintfhdl, u32 addr);
> +	u16 (*_read16)(struct intf_hdl *pintfhdl, u32 addr, int *error);
>  
>  	_read16 = pintfhdl->io_ops._read16;
>  
> -	r_val = _read16(pintfhdl, addr);
> +	r_val = _read16(pintfhdl, addr, error);
>  
>  	return r_val;
>  }

Same.

> -u32 _rtw_read32(struct adapter *adapter, u32 addr)
> +u32 _rtw_read32(struct adapter *adapter, u32 addr, int *error)
>  {
>  	u32 r_val;
>  	struct io_priv *pio_priv = &adapter->iopriv;
>  	struct	intf_hdl		*pintfhdl = &pio_priv->intf;
> -	u32	(*_read32)(struct intf_hdl *pintfhdl, u32 addr);
> +	u32	(*_read32)(struct intf_hdl *pintfhdl, u32 addr, int *error);
>  
>  	_read32 = pintfhdl->io_ops._read32;
>  
> -	r_val = _read32(pintfhdl, addr);
> +	r_val = _read32(pintfhdl, addr, error);
>  
>  	return r_val;
>  }

Same.

I'm done for now: too many lines to read all at once :)

Regards,

Fabio