On Mon, Jul 04, 2011 at 10:44:23AM +0200, Sebastian Andrzej Siewior wrote:
> Use proper printf length modifier for the dma_addr_t thingy. It is cast
> to u64 since dma_addr_t can be > 32bit on 32bit archs.
>
> |usb/gadget/net2272.c: In function 'net2272_kick_dma':
> |usb/gadget/net2272.c:740:2: warning: format '%08x' expects type 'long unsigned int', but argument 6 has type 'dma_addr_t'
> |usb/gadget/net2272.c: In function 'net2272_queue':
> |usb/gadget/net2272.c:859:2: warning: format '%08x' expects type 'unsigned int', but argument 8 has type 'dma_addr_t'
> |usb/gadget/ci13xxx_udc.c: In function 'show_registers':
> |usb/gadget/ci13xxx_udc.c:1242:1: warning: the frame size of 2064 bytes is larger than 1024 bytes
> |usb/gadget/fusb300_udc.c: In function â??fusb300_fill_idma_prdtblâ??:
> |usb/gadget/fusb300_udc.c:1074:12: warning: cast from pointer to integer of different size
> |usb/gadget/fusb300_udc.c: At top level:
> |usb/gadget/fusb300_udc.c:771:13: warning: â??fusb300_wrfifoâ?? defined but not used
> |usb/gadget/fusb300_udc.c:1027:13: warning: â??fusb300_set_ep_bycntâ?? defined but not used
> |usb/gadget/langwell_udc.c: In function 'queue_dtd':
> |usb/gadget/langwell_udc.c:596:79: warning: cast from pointer to integer of different size
> |usb/gadget/langwell_udc.c: In function 'langwell_ep_dequeue':
> |usb/gadget/langwell_udc.c:922:11: warning: cast from pointer to integer of different size
> |usb/gadget/langwell_udc.c: In function 'langwell_udc_probe':
> |usb/gadget/langwell_udc.c:3274:2: warning: format '%d' expects type 'int', but argument 4 has type 'size_t'
> |usb/gadget/langwell_udc.c:3289:2: warning: format '%d' expects type 'int', but argument 4 has type 'size_t'
> |usb/gadget/langwell_udc.c: In function 'langwell_udc_resume':
> |usb/gadget/langwell_udc.c:3473:2: warning: format '%d' expects type 'int', but argument 4 has type 'size_t'
> |usb/gadget/langwell_udc.c:3487:2: warning: format '%d' expects type 'int', but argument 4 has type 'size_t'
>
> fusb300_udc is a little more:
> - remove pointer u32 abuse in fusb300_fill_idma_prdtbl(). It is assigned
> the dma_addr to a pointer and then back. Poor families may have to
> recycle variables but we don't
> - don't free req.buf in error case. We don't do it in the ok case so it
> is probably wring to do it in error case.
> - return in error case. There is no reason to continue without data and
> performing ops on an invalid pointer.
> - The if (d) statement is bogus since an invalid DMA pointer is ~0 on
> some architecutres. And since we return for the invalid case we don't
> need it.
> - #if 0 unused code. If +2 cycles it is still there then I'm going to
> nuke it. If it is required due to broken DMA engine make an
> platform_data option.
>
> Both, langwell and fusb depend on not PHYS_ADDR_T_64BIT since they don't
> handle the larger addr space correctly. This should be fixed for them if
> they wish to work properly on bigmem kernels.
>
> Cc: Mike Frysinger <vapier@xxxxxxxxxx>
> Cc: Pavankumar Kondeti <pkondeti@xxxxxxxxxxxxxx>
> Cc: Yuan-Hsin Chen <yhchen@xxxxxxxxxxxxxxxx>
> Cc: Alan Cox <alan@xxxxxxxxxxxxxxx>
> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
NAK, too much in one patch :-(
> Greg, are all of them addressed?
>
> drivers/usb/gadget/Kconfig | 2 ++
> drivers/usb/gadget/ci13xxx_udc.c | 12 ++++++++++--
> drivers/usb/gadget/fusb300_udc.c | 26 +++++++++++---------------
> drivers/usb/gadget/langwell_udc.c | 12 ++++++------
> drivers/usb/gadget/net2272.c | 8 ++++----
> 5 files changed, 33 insertions(+), 27 deletions(-)
>
> diff --git a/drivers/usb/gadget/Kconfig b/drivers/usb/gadget/Kconfig
> index 46a253a..b0594d9 100644
> --- a/drivers/usb/gadget/Kconfig
> +++ b/drivers/usb/gadget/Kconfig
> @@ -157,6 +157,7 @@ config USB_FSL_USB2
>
> config USB_FUSB300
> tristate "Faraday FUSB300 USB Peripheral Controller"
> + depends on !PHYS_ADDR_T_64BIT
> select USB_GADGET_DUALSPEED
> help
> Faraday usb device controller FUSB300 driver
> @@ -425,6 +426,7 @@ config USB_GOKU
> config USB_LANGWELL
> tristate "Intel Langwell USB Device Controller"
> depends on PCI
> + depends on !PHYS_ADDR_T_64BIT
I would rather not. what if Intel comes up with a PCIe card to aid
development with this controller ?
> diff --git a/drivers/usb/gadget/fusb300_udc.c b/drivers/usb/gadget/fusb300_udc.c
> index 06353e7..277ebaa 100644
> --- a/drivers/usb/gadget/fusb300_udc.c
> +++ b/drivers/usb/gadget/fusb300_udc.c
> @@ -767,6 +767,7 @@ static void fusb300_rdfifo(struct fusb300_ep *ep,
> } while (!reg);
> }
>
> +#if 0
> /* write data to fifo */
> static void fusb300_wrfifo(struct fusb300_ep *ep,
> struct fusb300_request *req)
> @@ -816,6 +817,7 @@ static void fusb300_wrfifo(struct fusb300_ep *ep,
> i++;
> } while (!reg);
> }
> +#endif
>
> static u8 fusb300_get_epnstall(struct fusb300 *fusb300, u8 ep)
> {
> @@ -1024,6 +1026,7 @@ static int setup_packet(struct fusb300 *fusb300, struct usb_ctrlrequest *ctrl)
> return ret;
> }
>
> +#if 0
> static void fusb300_set_ep_bycnt(struct fusb300_ep *ep, u32 bycnt)
> {
> struct fusb300 *fusb300 = ep->fusb300;
> @@ -1034,6 +1037,7 @@ static void fusb300_set_ep_bycnt(struct fusb300_ep *ep, u32 bycnt)
>
> iowrite32(reg, fusb300->reg + FUSB300_OFFSET_EPFFR(ep->epnum));
> }
> +#endif
I would rather remove the dead code. I'm sure there are people keeping
local patches to invert the if 0 crap so they can test PIO mode. It
would better if they would send us a proper patch making that runtime
selectable.
--
balbi
Attachment:
signature.asc
Description: Digital signature
