Re: [PATCH] 2/7 AU1100 MMC support

[Sergei Shtylyov <sshtylyov@ru.mvista.com>]

Hello.

Rodolfo Giometti wrote:

>>>static void au1xmmc_set_power(struct au1xmmc_host *host, int state)
>>>{

>>>-	u32 val = au1xmmc_card_table[host->id].bcsrpwr;
>>>+	u32 val;

>>>+	val = au1xmmc_card_table[host->id].power;
>>>+
>>>+#if defined(CONFIG_MIPS_DB1200)
>>>	bcsr->board &= ~val;
>>>	if (state) bcsr->board |= val;
>>>+#endif

>>>	au_sync_delay(1);
>>>}

>>   If DBAu1100 doesn't allow to control slot power, then I don't think 
>>pretending it does is a good thing. Shouldn't these #ifdef's be in 
>>au1xmmc_set_ios() instead (the function is void anyway but that would allow 
>>us to save on the code size a bit more)?

> Mmm. I proposed that solution but I don't know exaclty how several
> DB1x00 boards work. I just protect the variable "bcsr" which is not
> defined for my board.

    That wasn't obvious from the patch that this is intended for some specific 
board and BCSRs (board control/status registers) seem to exist on all Alchemy 
development boards.

>>>static inline int au1xmmc_card_inserted(struct au1xmmc_host *host)
>>>{
>>>-	return (bcsr->sig_status & au1xmmc_card_table[host->id].bcsrstatus)
>>>-		? 1 : 0;
>>>+	u32 val, data = 1;
>>>+
>>>+	val = au1xmmc_card_table[host->id].status;
>>>+
>>>+#if defined(CONFIG_MIPS_DB1200)
>>>+	data = bcsr->sig_status & val;
>>>+#endif
>>>+
>>>+	return !!data;
>>>}

>>   Hrm, are you sure there's no way to sense that the card is *really* 
>>inserted or not?
> 
> 
> Again as above. For my specific board I use:

>    #if defined(CONFIG_MIPS_DB1200)
>        data = bcsr->sig_status & val;
>    #elif defined(CONFIG_MIPS_MYBOARD)
>        specific code...
>    #endif

> Maybe we should modify my solution including other DB1x00 boards
> ifdef. However, the important thing is to protect againt variable
> "bcsr" if a specific board doesn't support it.

    If you're not introducing the support for the new board, introducing those 
#ifdef's doesn't make much sense.
    Overall, the support for the boards other than Pb/DbAu1200 seems broken in 
that driver -- see below why...

>>>static inline int au1xmmc_card_readonly(struct au1xmmc_host *host)
>>>{
>>>-	return (bcsr->status & au1xmmc_card_table[host->id].wpstatus)
>>>-		? 1 : 0;
>>>+	u32 val, data = 0;
>>>+
>>>+	val = au1xmmc_card_table[host->id].wpstatus;
>>>+
>>>+#if defined(CONFIG_MIPS_DB1200)
>>>+	data = bcsr->status & val;
>>>+#endif
>>>+
>>>+	return !!data;
>>>}

>>   Ditto.

> The same as above. :)

    Indeed, some modifications are needed for the other Alchemy boards -- 
there's some discrepancy with the "board" register in particular, for the 
boards older than Pb/DBAu1200 it was called "specific" having much the same 
format and purpose. This is not so however with "sig_status" and "status" 
registers used to sense card's presence and for write protecting. Judging on 
<asm/mach-db1x00/db1x00.h>, "specific" reg. should be used instead. There are 
MMC macros in that file BTW for detecting the cards and applying power to them.

> Ciao,

WBR, Sergei