Hi Uffe, > -----Original Message----- > From: Ulf Hansson [mailto:ulf.hansson@xxxxxxxxxx] > Sent: 2017年12月23日 23:24 > To: Y.b. Lu <yangbo.lu@xxxxxxx> > Cc: Yinbo Zhu <yinbo.zhu@xxxxxxx>; Adrian Hunter > <adrian.hunter@xxxxxxxxx>; linux-mmc@xxxxxxxxxxxxxxx > Subject: Re: [PATCH v1] mmc: sdhci-of-esdhc: Workaround for reducing the > maximum speed on ls1021atwr > > On 22 December 2017 at 09:13, Y.b. Lu <yangbo.lu@xxxxxxx> wrote: > > Hi Uffe, > > > > > >> -----Original Message----- > >> From: Ulf Hansson [mailto:ulf.hansson@xxxxxxxxxx] > >> Sent: 2017年11月16日 15:34 > >> To: Yinbo Zhu <yinbo.zhu@xxxxxxx> > >> Cc: Adrian Hunter <adrian.hunter@xxxxxxxxx>; Y.b. Lu > >> <yangbo.lu@xxxxxxx>; linux-mmc@xxxxxxxxxxxxxxx > >> Subject: Re: [PATCH v1] mmc: sdhci-of-esdhc: Workaround for reducing > >> the maximum speed on ls1021atwr > >> > >> On 15 November 2017 at 10:01, <yinbo.zhu@xxxxxxx> wrote: > >> > From: "yinbo.zhu" <yinbo.zhu@xxxxxxx> > >> > > >> > In SDHC high speed AC timing, the tshivkh parameter is defined as > >> > input setup times:SDHC_CMD, SDHC_DATx, to SDHC_CLK. The value of > >> > the tshivkh should be 2.5 ns considering the round trip delay, > >> > board/data skew. > >> > However, because of this erratum, it needs at least 4.1 ns. > >> > > >> > eSDHC cannot run at the maximum clock speed for the high speed > >> > mode, or there is a limit on the length of the trace on the board > >> > for data, command, and clock lines of the SDHC. > >> > > >> > Signed-off-by: yinbo.zhu <yinbo.zhu@xxxxxxx> > >> > --- > >> > drivers/mmc/host/sdhci-of-esdhc.c | 6 ++++++ > >> > 1 file changed, 6 insertions(+) > >> > > >> > diff --git a/drivers/mmc/host/sdhci-of-esdhc.c > >> > b/drivers/mmc/host/sdhci-of-esdhc.c > >> > index 023c24bd0d94..2744dd58a573 100644 > >> > --- a/drivers/mmc/host/sdhci-of-esdhc.c > >> > +++ b/drivers/mmc/host/sdhci-of-esdhc.c > >> > @@ -498,6 +498,12 @@ static void esdhc_of_set_clock(struct > >> > sdhci_host > >> *host, unsigned int clock) > >> > clock -= 5000000; > >> > } > >> > > >> > + /* Workaround to reduce the clock frequency for ls1021a esdhc > */ > >> > + if (of_find_compatible_node(NULL, NULL, > >> > + "fsl,ls1021a-esdhc")) { > >> > >> It's better to use the ->data pointer in the struct of_device_id for > >> this kind of variant data. > >> > >> In ->probe() you then call of_match_device() and pick up the ->data > >> pointer and assign it to your driver private data. > >> > >> Many drivers already do like this. Have a look at dw_mmc-exynos.c for > >> example. > >> > > > > [Y.b. Lu] How about using soc_device_match()? There is also .data member > could be used in soc_device_attribute structure. > > Use soc_device_match() in esdhc_init and get .data information into > sdhci_esdhc structure which is private data. > > Current incorrect host version register value fix-up for some platforms is > using this method. > > I think you need to answer the question: is the problem SoC specific or > device-variant specific. Depending on the answer, you should pick the correct > solution. > [Y.b. Lu] This should be SoC specific. It's silicon errata. It has possibility to be fixed in future silicon revision. So soc_device_match() could be used I think :) > [...] > > Kind regards > Uffe ��.n��������+%������w��{.n�����{��i��)�� jg��� �����ݢj����G���� ���j:+v���w�m������w�������h�����٥
