Re: [PATCH RFC v2 0/6] staging: r8188eu: avoid uninit value bugs

[Phillip Potter]

On Sun, 22 Aug 2021 at 15:35, Pavel Skripkin <paskripkin@xxxxxxxxx> wrote:
>
> Hi, Greg, Larry and Phillip!
>
> I noticed, that new staging driver was added like 3 weeks ago and I decided
> to look at the code, because drivers in staging directory are always buggy.
>
> The first thing I noticed is *no one* was checking read operations result, but
> it can fail and driver may start writing random stack values into registers. It
> can cause driver misbehavior or device misbehavior.
>
> To avoid this type of bugs, i've changed rtw_read* API. Now all rtw_read
> funtions return an error, when something went wrong with usb transfer.
>
> It helps callers to break/return earlier and don't write random values to
> registers or to rely on random values.
>
> Why is this pacth series RFC?
>   1. I don't have this device and I cannot test these changes.
>   2. I don't know how to handle errors in each particular case. For now, function
>      just returns or returns an error. That's all. I hope, driver maintainers will
>      help with these bits.
>   3. I guess, I handled not all uninit value bugs here. I hope, I fixed
>      at least half of them
>
>
> v1 -> v2:
>   1. Make rtw_read*() return an error instead of initializing pointer to error
>   2. Split one huge patch to smaller ones for each rtw_read{8,16,32} function
>      changes
>   3. Add new macro for printing register values (It helps to not copy-paste error
>      handling)
>   4. Removed {read,write}_macreg (Suggested by Phillip)
>   5. Rebased on top of staging-next
>   6. Cleaned checkpatch errors and warnings
>
> Only build-tested, since I don't have device with r8118eu chip
>
> Pavel Skripkin (6):
>   staging: r8188eu: remove {read,write}_macreg
>   staging: r8188eu: add helper macro for printing registers
>   staging: r8188eu: add error handling of rtw_read8
>   staging: r8188eu: add error handling of rtw_read16
>   staging: r8188eu: add error handling of rtw_read32
>   staging: r8188eu: make ReadEFuse return an int
>
>  drivers/staging/r8188eu/core/rtw_debug.c      |  79 +++-
>  drivers/staging/r8188eu/core/rtw_efuse.c      | 125 +++--
>  drivers/staging/r8188eu/core/rtw_io.c         |  27 +-
>  drivers/staging/r8188eu/core/rtw_mp.c         |  70 ++-
>  drivers/staging/r8188eu/core/rtw_mp_ioctl.c   |  13 +-
>  drivers/staging/r8188eu/core/rtw_pwrctrl.c    |   5 +-
>  drivers/staging/r8188eu/core/rtw_sreset.c     |   9 +-
>  .../r8188eu/hal/Hal8188ERateAdaptive.c        |   8 +-
>  drivers/staging/r8188eu/hal/HalPhyRf_8188e.c  |  21 +-
>  drivers/staging/r8188eu/hal/HalPwrSeqCmd.c    |   9 +-
>  drivers/staging/r8188eu/hal/hal_com.c         |  23 +-
>  drivers/staging/r8188eu/hal/hal_intf.c        |   6 +-
>  drivers/staging/r8188eu/hal/odm_interface.c   |  12 +-
>  drivers/staging/r8188eu/hal/rtl8188e_cmd.c    |  33 +-
>  drivers/staging/r8188eu/hal/rtl8188e_dm.c     |   6 +-
>  .../staging/r8188eu/hal/rtl8188e_hal_init.c   | 285 +++++++++---
>  drivers/staging/r8188eu/hal/rtl8188e_phycfg.c |  27 +-
>  drivers/staging/r8188eu/hal/rtl8188e_sreset.c |  22 +-
>  drivers/staging/r8188eu/hal/rtl8188eu_led.c   |  18 +-
>  drivers/staging/r8188eu/hal/usb_halinit.c     | 439 +++++++++++++++---
>  drivers/staging/r8188eu/hal/usb_ops_linux.c   |  57 ++-
>  drivers/staging/r8188eu/include/hal_intf.h    |   6 +-
>  .../staging/r8188eu/include/odm_interface.h   |   6 +-
>  .../staging/r8188eu/include/rtl8188e_hal.h    |   2 +-
>  drivers/staging/r8188eu/include/rtw_debug.h   |  13 +
>  drivers/staging/r8188eu/include/rtw_efuse.h   |   4 +-
>  drivers/staging/r8188eu/include/rtw_io.h      |  18 +-
>  drivers/staging/r8188eu/include/rtw_mp.h      |   2 -
>  drivers/staging/r8188eu/os_dep/ioctl_linux.c  | 179 +++++--
>  drivers/staging/r8188eu/os_dep/usb_intf.c     |   3 +-
>  30 files changed, 1138 insertions(+), 389 deletions(-)
>
> --
> 2.32.0
>

Dear Pavel,

Thanks for this. I like the code a lot. One thing I am conflicted on
is the helper macro for the printing of register values though. Whilst
I'm not necessarily opposed to the concept of the macro itself, I
don't think it should rely on GlobalDebugLevel for one thing - if we
are going to control printing of messages at runtime then in my mind
this should be done via debugfs and pr_debug or similar - an in-kernel
mechanism rather than something driver-provided. Also, the example you
give of:

        u32 tmp;
        if (!rtw_read(&tmp))
                DBG("reg = %d\n", tmp);

Doesn't seem overly unclear to me if DBG was a pr_debug or similar,
but I get what you're saying about repetition. This is just a small
thing though, would be interested to see what others think. Many
thanks.

Regards,
Phil