[RFC][PATCH V3] axi: add AXI bus driver

[Rafał Miłecki]

2011/4/12 Joe Perches <joe at perches.com>:
> On Mon, 2011-04-11 at 23:25 +0200, Rafał Miłecki wrote:
>
>>   Proper pr_* usage
>
> Just trivia...

The patch you were reviewing does not have pr_* stuff included, sorry.
I published V3 by mistake missing some changes. Patch V3.1 alreay
fixes that.


>> diff --git a/drivers/axi/axi_private.h b/drivers/axi/axi_private.h
>> new file mode 100644
>> index 0000000..1c995f5
>> --- /dev/null
>> +++ b/drivers/axi/axi_private.h
>> @@ -0,0 +1,45 @@
>> +#ifndef LINUX_AXI_PRIVATE_H_
>> +#define LINUX_AXI_PRIVATE_H_
>> +
>> +#include <linux/axi/axi.h>
>> +#include <linux/device.h>
>> +#include <linux/list.h>
>> +#include <linux/types.h>
>> +#include <linux/spinlock.h>
>> +#include <linux/pci.h>
>> +#include <linux/mod_devicetable.h>
>> +#include <linux/dma-mapping.h>
>> +#include <linux/types.h>
>
> Duplicate include of linux/types.h

V3.1


>> +
>> +#ifndef pr_fmt
>> +#define pr_fmt(fmt)          "axi: " fmt
>> +#endif
>
> This needs to be before any include that includes kernel.h
> and I believe
> #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
> is the same output and more normally used style.
>
> Right now, you've got no prefix for any format string.
> Try '$ strings built-in.o | grep "^<.>"' and see.

V3.1

>> +void axi_pmu_workarounds(struct axi_drv_cc *cc)
>> +{
>> +     struct axi_bus *bus = cc->core->bus;
>> +
>> +     switch (bus->chipinfo.id) {
>> +     case 0x4313:
>
> Probably better to use defines.

That magic hex values always match common name of chipsets, so it's
looks pretty clean/obvious... Not sure if it's worth it?

>> +             axi_chipco_chipctl_maskset(cc, 0, ~0, 0x7);
>> +             break;
>> +     case 0x4331:
>> +             axi_err("Enabling Ext PA lines not implemented\n");
>
> You should still remove axi_err/info/debug and
> just use pr_<level>.

V3.1

>> +             break;
>> +     case 43224:
>> +             if (bus->chipinfo.rev == 0) {
>> +                     axi_err("Workarounds for 43224 rev 0 not fully "
>> +                                     "implemented\n");
>
> I suggest you don't bother splitting and wrapping
> format strings > 80 chars.  Splitting makes grep
> a bit harder.
>
>                        pr_err("Workarounds for 43224 rev 0 not fully implemented\n");
>
> Some hex and some decimal cases?
> Maybe some comments for that?

Well, such a trick from Broadcom. I believe wanted to keep 43* names
for the chipsets, they invented sth like 43224 but they had only
0xFFFF. So the used 0xA8D8==43224.

-- 
Rafał