[U-Boot] [PATCH 04/11] net: macb: Fix clk API usage for RISC-V systems

[Anup Patel]

> -----Original Message-----
> From: Alexander Graf [mailto:agraf at suse.de]
> Sent: Friday, January 18, 2019 5:05 PM
> To: Anup Patel <Anup.Patel at wdc.com>
> Cc: Rick Chen <rick at andestech.com>; Bin Meng <bmeng.cn at gmail.com>;
> Joe Hershberger <joe.hershberger at ni.com>; Lukas Auer
> <lukas.auer at aisec.fraunhofer.de>; Masahiro Yamada
> <yamada.masahiro at socionext.com>; Simon Glass <sjg at chromium.org>;
> Palmer Dabbelt <palmer at sifive.com>; Paul Walmsley
> <paul.walmsley at sifive.com>; Atish Patra <Atish.Patra at wdc.com>;
> Christoph Hellwig <hch at infradead.org>; U-Boot Mailing List <u-
> boot at lists.denx.de>
> Subject: Re: [PATCH 04/11] net: macb: Fix clk API usage for RISC-V systems
> 
> 
> 
> > Am 18.01.2019 um 07:05 schrieb Anup Patel <Anup.Patel at wdc.com>:
> >
> >
> >
> >> -----Original Message-----
> >> From: Alexander Graf [mailto:agraf at suse.de]
> >> Sent: Thursday, January 17, 2019 11:37 PM
> >> To: Anup Patel <Anup.Patel at wdc.com>; Rick Chen
> <rick at andestech.com>;
> >> Bin Meng <bmeng.cn at gmail.com>; Joe Hershberger
> >> <joe.hershberger at ni.com>; Lukas Auer
> >> <lukas.auer at aisec.fraunhofer.de>; Masahiro Yamada
> >> <yamada.masahiro at socionext.com>; Simon Glass <sjg at chromium.org>
> >> Cc: Palmer Dabbelt <palmer at sifive.com>; Paul Walmsley
> >> <paul.walmsley at sifive.com>; Atish Patra <Atish.Patra at wdc.com>;
> >> Christoph Hellwig <hch at infradead.org>; U-Boot Mailing List <u-
> >> boot at lists.denx.de>
> >> Subject: Re: [PATCH 04/11] net: macb: Fix clk API usage for RISC-V
> >> systems
> >>
> >>> On 01/17/2019 11:38 AM, Anup Patel wrote:
> >>> This patch does following fixes in MACB ethernet driver for using it
> >>> on RISC-V systems (particularly QEMU sifive_u
> >>> machine):
> >>> 1. asm/arch/clk.h is not available on RISC-V port so include
> >>>    it only for non-RISC-V systems.
> >>> 2. Don't fail in macb_enable_clk() if clk_enable() returns
> >>>    -ENOSYS because we get -ENOSYS for fixed-rate clocks.
> >>>
> >>> Signed-off-by: Anup Patel <anup.patel at wdc.com>
> >>> Reviewed-by: Bin Meng <bmeng.cn at gmail.com>
> >>> ---
> >>>  drivers/net/macb.c | 4 +++-
> >>>  1 file changed, 3 insertions(+), 1 deletion(-)
> >>>
> >>> diff --git a/drivers/net/macb.c b/drivers/net/macb.c index
> >>> 94c89c762b..9a06b523cc 100644
> >>> --- a/drivers/net/macb.c
> >>> +++ b/drivers/net/macb.c
> >>> @@ -38,7 +38,9 @@
> >>>  #include <linux/mii.h>
> >>>  #include <asm/io.h>
> >>>  #include <asm/dma-mapping.h>
> >>> +#ifndef CONFIG_RISCV
> >>>  #include <asm/arch/clk.h>
> >>> +#endif
> >>>  #include <linux/errno.h>
> >>>
> >>>  #include "macb.h"
> >>> @@ -1066,7 +1068,7 @@ static int macb_enable_clk(struct udevice
> *dev)
> >>>       */
> >>>  #ifndef CONFIG_MACB_ZYNQ
> >>>      ret = clk_enable(&clk);
> >>
> >> If clk.h is not available, who exports clk_enable() then; and why is
> >> the included needed in the first place?
> >
> > For some of the ARM boards, clk instances are provided directly by
> > arch/arm/mach-xyz sources. For such boards, asm/arch/clk.h is
> > required. I think these boards should move to DT based clk drivers.
> 
> Can you at least make this a positive #ifdef then rather than ifndef? We want
> to isolate the odd case, not the normal one.
> 
> Maybe you can even find a config option that isolates it further?

With the additional check "ret != -ENOSYS", we don't require the #ifndef but
someone with Zynq SOC board need to confirm that removing #ifndef won't
hurt.

Regards,
Anup