lkml.org 
[lkml]   [2016]   [Jun]   [17]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    Date
    From
    SubjectRe: [PATCH] mmc: dw_mmc: remove UBSAN warning in dw_mci_setup_bus()
    Hello Jaehoon,

    On 2016년 06월 17일 10:30, Jaehoon Chung wrote:
    > Hi Seung-Woo,
    >
    > On 06/10/2016 10:29 AM, Seung-Woo Kim wrote:
    >> Hi Jaehoon,
    >>
    >> On 2016년 06월 09일 21:38, Jaehoon Chung wrote:
    >>> Hi Seung-Woo,
    >>>
    >>> On 06/08/2016 01:07 PM, Seung-Woo Kim wrote:
    >>>> This patch removes following UBSAN warnings in dw_mci_setup_bus().
    >>>> The warnings are caused because of shift with more than 31 on 32
    >>>> bit variable, so this patch fixes to shift only for less than 32.
    >>>>

    <snip.>

    >>>>
    >>>> diff --git a/drivers/mmc/host/dw_mmc.c b/drivers/mmc/host/dw_mmc.c
    >>>> index 2cc6123..dff045e 100644
    >>>> --- a/drivers/mmc/host/dw_mmc.c
    >>>> +++ b/drivers/mmc/host/dw_mmc.c
    >>>> @@ -1099,7 +1099,8 @@ static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
    >>>>
    >>>> div = (host->bus_hz != clock) ? DIV_ROUND_UP(div, 2) : 0;
    >>>>
    >>>> - if ((clock << div) != slot->__clk_old || force_clkinit)
    >>>> + if (((div < 32) ? (clock << div) : 0) != slot->__clk_old ||
    >>>
    >>> Well, we don't expect that clock is 0.
    >>> if clock is 0, it should be passed to " if (!clock)"..
    >>>
    >>> During initializing card, clock is 400KHz or less..I understood what you want to fix.
    >>> But I taught this is not correct.
    >>
    >> It seems slot->__clk_old is not really clock but a kind of value to
    >> store both clock and div in one variable. And at least, my compiler
    >> calculates right shift with more than 32 as 0. I am not sure there is
    >> other proper value for the case.
    >>
    >> By the way, in my test environment, clock is calculated as like
    >> following steps:
    >> mmc_host mmc0: Bus speed (slot 0) = 200000000Hz (slot req 400000Hz,
    >> actual 400000HZ div = 250)
    >> mmc_host mmc0: Bus speed (slot 0) = 200000000Hz (slot req 200000000Hz,
    >> actual 200000000HZ div = 0)
    >> mmc_host mmc0: Bus speed (slot 0) = 200000000Hz (slot req 52000000Hz,
    >> actual 50000000HZ div = 2)
    >> mmc_host mmc0: Bus speed (slot 0) = 400000000Hz (slot req 52000000Hz,
    >> actual 50000000HZ div = 4)
    >> mmc_host mmc0: Bus speed (slot 0) = 400000000Hz (slot req 200000000Hz,
    >> actual 200000000HZ div = 1)
    >>
    >> and only the first case is reported as the warning. If you let me know
    >> any proper value, then I will fix with the value.
    >
    > I think that "clock << div" and__old_clock are used for checking whether clock is changed.
    > So it doesn't need to rotate with div.
    >
    > Could you check the below codes?
    >
    > if ((clock != slot->clk_old) || force_clkinit)
    > ...
    >
    > slot->__clk_old = clock;

    From the comment about alignment of slot->__clk_old, it says that "keep
    the clock with reflecting clock dividor". So to me, it is better also
    adding slot->__dlv with your suggestion. I will send patch like it, soon.

    Regards,
    - Seung-Woo Kim

    >
    > Best Regards,
    > Jaehoon Chung
    >
    >>
    >> Thanks,
    >> - Seung-Woo Kim
    >>
    >>
    >>>
    >>>
    >>>> + force_clkinit)
    >>>> dev_info(&slot->mmc->class_dev,
    >>>> "Bus speed (slot %d) = %dHz (slot req %dHz, actual %dHZ div = %d)\n",
    >>>> slot->id, host->bus_hz, clock,
    >>>> @@ -1129,7 +1130,7 @@ static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
    >>>> mci_send_cmd(slot, sdmmc_cmd_bits, 0);
    >>>>
    >>>> /* keep the clock with reflecting clock dividor */
    >>>> - slot->__clk_old = clock << div;
    >>>> + slot->__clk_old = (div < 32) ? (clock << div) : 0;

    <snip.>

    --
    Seung-Woo Kim
    Samsung Software R&D Center
    --

    \
     
     \ /
      Last update: 2016-06-17 06:41    [W:3.733 / U:1.040 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site