linux-kernel - Re: [PATCH 2/3] sdhci: sparx5: Add Sparx5 SoC eMMC driver

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:   Mon, 25 May 2020 16:26:59 +0200
From:   Lars Povlsen <lars.povlsen@...rochip.com>
To:     Adrian Hunter <adrian.hunter@...el.com>
CC:     Lars Povlsen <lars.povlsen@...rochip.com>,
        Ulf Hansson <ulf.hansson@...aro.org>,
        SoC Team <soc@...nel.org>,
        "Microchip Linux Driver Support" <UNGLinuxDriver@...rochip.com>,
        <linux-mmc@...r.kernel.org>, <devicetree@...r.kernel.org>,
        <linux-arm-kernel@...ts.infradead.org>,
        <linux-kernel@...r.kernel.org>,
        Alexandre Belloni <alexandre.belloni@...tlin.com>
Subject: Re: [PATCH 2/3] sdhci: sparx5: Add Sparx5 SoC eMMC driver


Adrian Hunter writes:

> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
>
> On 20/05/20 2:14 pm, Lars Povlsen wrote:
>>
>> Lars Povlsen writes:
>>
>>> Adrian Hunter writes:
>>>
>>>> On 13/05/20 4:31 pm, Lars Povlsen wrote:
>>>>> This adds the eMMC driver for the Sparx5 SoC. It is based upon the
>>>>> designware IP, but requires some extra initialization and quirks.
>>>>>
>>>>> Reviewed-by: Alexandre Belloni <alexandre.belloni@...tlin.com>
>>>>> Signed-off-by: Lars Povlsen <lars.povlsen@...rochip.com>
>>>>> ---
>> {Snip]
>>>>> +};
>>>>> +
>>>>> +static const struct sdhci_pltfm_data sdhci_sparx5_pdata = {
>>>>> +     .quirks  = 0,
>>>>> +     .quirks2 = SDHCI_QUIRK2_HOST_NO_CMD23 | /* Card quirk */
>>>>
>>>> If this is a card quirk then it should be in drivers/mmc/core/quirks.h not here.
>>>
>>
>> Adrian, I had a go at changing the controller quirk to a card quirk.
>>
>> Unfortunately, SDHCI_QUIRK2_HOST_NO_CMD23 does not directly translate to
>> MMC_QUIRK_BLK_NO_CMD23, as for 'do_rel_wr' in mmc_blk_rw_rq_prep(), it
>> will *still* use MMC_SET_BLOCK_COUNT (cmd23), causing the issue.
>>
>> We are using a ISSI "IS004G" device, and so I have gone through the
>> motions of adding it to quirks.h. The comment before the list of devices
>> using MMC_QUIRK_BLK_NO_CMD23 suggest working around a performance issue,
>> which is not exactly the issue I'm seeing. I'm seeing combinations of
>> CMD_TOUT_ERR, DATA_CRC_ERR and DATA_END_BIT_ERR whenever a cmd23 is
>> issued.
>>
>> I have not been able to test the controller with another eMMC device
>> yet, but I expect its not the controller at fault.
>>
>> So, I'm a little bit in doubt of how to proceed - either keep the quirk
>> as a controller quirk - or make a *new* card quirk (with
>> SDHCI_QUIRK2_HOST_NO_CMD23 semantics)?
>>
>> Anybody else have had experience with ISSI eMMC devices?
>>
>> I have also tried to use DT sdhci-caps-mask, but MMC_CAP_CMD23 is not
>> read from the controller just (unconditionally) set in sdhci.c - so that
>> doesn't fly either.
>>
>> Any suggestions?
>
> It is up to you.  In the future, you may want to distinguish devices that
> have this problem from ones that do not.
>
> If you are not sure it is the ISSI eMMC, and maybe not the host controller,
> then might it be the board?  Perhaps make SDHCI_QUIRK2_HOST_NO_CMD23
> conditional on the particular compatibility string?
>
> At a minimum, change the "/* Card quirk */" comment to a fuller explanation.
>

Adrian, I'm getting a board ready with another eMMC device, and we're
also trying to contact ISSI for info.

My hope is to at least verify whether this is a controller or a card
issue one way or the other. Then, I'll choose an appropriate solution
for it.

Thank you for your advice so far.

---Lars

>>
>>> Yes, its supposedly a card quirk. I'll see to use the card quirks
>>> methods in place.
>>>
>>

-- 
Lars Povlsen,
Microchip