From: =?ISO-8859-1?Q?Ren=E9_Griessl?= Date: Wed, 10 Sep 2014 12:00:29 +0200 Subject: [U-Boot] [PATCH v2 1/2] usb: eth: add ASIX AX88179 DRIVER In-Reply-To: <201409091634.19271.marex@denx.de> References: <1409754681-12855-1-git-send-email-rgriessl@cit-ec.uni-bielefeld.de> <201409091634.19271.marex@denx.de> Message-ID: <5410213D.90805@cit-ec.uni-bielefeld.de> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de Am 09.09.2014 16:34, schrieb Marek Vasut: > On Wednesday, September 03, 2014 at 04:31:20 PM, Rene Griessl wrote: >> changes in v2: >> -added usb_ether.h to change list >> -added 2nd patch to enable driver for arndale board >> >> Signed-off-by: Rene Griessl > I see that in Linux, there is asix_common.c stuff. Can this driver share any > parts with drivers/net/ax88* ? The asix_common.c includes asix.h. There you see that the common driver supports following devices: AX88172 AX88772 AX88178 but it is not supporting AX88179 and AX88178a, they have a separate driver called ax88179_178a.c These two have a different style in accessing MAC registers and PHY >> --- >> drivers/usb/eth/Makefile | 3 + >> drivers/usb/eth/asix88179.c | 641 >> ++++++++++++++++++++++++++++++++++++++++++++ drivers/usb/eth/usb_ether.c | >> 7 + >> include/usb_ether.h | 6 + >> 4 files changed, 657 insertions(+) >> create mode 100644 drivers/usb/eth/asix88179.c >> >> diff --git a/drivers/usb/eth/Makefile b/drivers/usb/eth/Makefile >> index 94551c4..fad4acd 100644 >> --- a/drivers/usb/eth/Makefile >> +++ b/drivers/usb/eth/Makefile >> @@ -8,5 +8,8 @@ obj-$(CONFIG_USB_HOST_ETHER) += usb_ether.o >> ifdef CONFIG_USB_ETHER_ASIX >> obj-y += asix.o >> endif >> +ifdef CONFIG_USB_ETHER_ASIX_88179 >> +obj-y += asix88179.o >> +endif > This should be obj-$(CONFIG....) as seen below. Fix the asix one in a separate > patch please. > > [...] OK >> +/* ASIX AX88179 based USB 3.0 Ethernet Devices */ >> +#define AX88179_PHY_ID 0x03 >> +#define AX_EEPROM_LEN 0x100 >> +#define AX88179_EEPROM_MAGIC 0x17900b95 >> +#define AX_MCAST_FLTSIZE 8 >> +#define AX_MAX_MCAST 64 >> +#define AX_INT_PPLS_LINK ((u32)BIT(16)) > The u32 cast is not needed. Also, please drop the BIT() macro, it's just > obfuscating the code, just use (1 << 16) instead. Please fix globally. OK (was just copy'n'paste from the linux driver) >> +#define AX_RXHDR_L4_TYPE_MASK 0x1c >> +#define AX_RXHDR_L4_TYPE_UDP 4 >> +#define AX_RXHDR_L4_TYPE_TCP 16 >> +#define AX_RXHDR_L3CSUM_ERR 2 >> +#define AX_RXHDR_L4CSUM_ERR 1 >> +#define AX_RXHDR_CRC_ERR ((u32)BIT(29)) >> +#define AX_RXHDR_DROP_ERR ((u32)BIT(31)) >> +#define AX_ACCESS_MAC 0x01 >> +#define AX_ACCESS_PHY 0x02 >> +#define AX_ACCESS_EEPROM 0x04 >> +#define AX_ACCESS_EFUS 0x05 >> +#define AX_PAUSE_WATERLVL_HIGH 0x54 >> +#define AX_PAUSE_WATERLVL_LOW 0x55 > [...] > >> +static inline int asix_get_phy_addr(struct ueth_data *dev) >> +{ >> + ALLOC_CACHE_ALIGN_BUFFER(u8, buf, 2); >> + int ret = -1; >> + if (dev->pusb_dev->descriptor.idProduct == 0x1790) { >> + ret = asix_read_cmd(dev, AX_ACCESS_MAC, 0x10, 6, 6, buf); >> + debug("asix_get_phy_addr() returning > 0x%02x%02x%02x%02x%02x%02x\n", >> + buf[0], buf[1], buf[2], buf[3], buf[4], buf[5]); >> + } >> + else { >> + } > Uh, this check needs some rework, right ? Also, you want to lint your patches > with ./scripts/checkpatch.pl tool before resubmitting. was OK for ./scripts/checkpatch.pl but I can change that >> + return ret; >> +} >> + >> + >> +static int asix_read_mac(struct eth_device *eth) >> +{ >> + struct ueth_data *dev = (struct ueth_data *)eth->priv; >> + ALLOC_CACHE_ALIGN_BUFFER(unsigned char, buf, ETH_ALEN); >> + >> + if (dev->pusb_dev->descriptor.idProduct == 0x1790) { >> + asix_read_cmd(dev, AX_ACCESS_MAC, 0x10, 6, 6, buf); >> + debug("asix_read_mac() returning 0x%02x%02x%02x%02x%02x%02x\n", >> + buf[0], buf[1], buf[2], buf[3], buf[4], buf[5]); >> + memcpy(eth->enetaddr, buf, ETH_ALEN); >> + } >> + return 0; >> +} >> + >> + >> + >> +static int asix_basic_reset(struct ueth_data *dev) >> +{ >> + ALLOC_CACHE_ALIGN_BUFFER(u8, buf, 6); > Why does the buffer need to be aligned here ? It's just a buffer that is not > used for DMA, no ? > >> + u16 *tmp16; > Is it because you were getting unaligned accesses , since when the buffer itself > was not aligned and you did cast it to u16, the CPU triggered unaligned access ? Thats right, if I do not align I get unaligned accesses during USB communication. Is there a smarter solution for that? >> + u8 *tmp; > [...]