When network protocol errors occur (such as a file not being found on a
TFTP server), the processing done by the NetReceive() function will end
up calling the driver's .halt() implementation. However, after that the
device no longer has access to the memory buffers and will cause errors
such as this in the rtl_recv() function when trying to hand descriptors
back to the device:
pci_hose_bus_to_phys: invalid physical address
This can be fixed by deferring processing of network packets until the
descriptors have been handed back. That way rtl_halt() tearing down
network buffers is not going to prevent access to the buffers.
Reported-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Use size_t type for positive offsets instead of the loff_t type. The
later is defined as long long, which is larger than the pointer type
on OpenRISC architecture and therefore the following warning was
generated:
"warning: cast to pointer from integer of different size"
Signed-off-by: Vasili Galka <vvv444@gmail.com>
When testing the FEC driver on a mx6solox we noticed that the TDAR bit gets
always cleared prior then the READY bit is cleared in the last BD, which causes
FEC packets reception to always fail.
As explained by Ye Li:
"The TDAR bit is cleared when the descriptors are all out from TX ring, but on
mx6solox we noticed that the READY bit is still not cleared right after TDAR.
These are two distinct signals, and in IC simulation, we found that TDAR always
gets cleared prior than the READY bit of last BD becomes cleared.
In mx6solox, we use a later version of FEC IP. It looks like that this
intrinsic behaviour of TDAR bit has changed in this newer FEC version."
Fix this by polling the READY bit of BD after the TDAR polling, which covers the
mx6solox case and does not harm the other SoCs.
No performance drop has been noticed with this patch applied when testing TFTP
transfers on several boards of different i.mx SoCs.
Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Acked-by: Marek Vasut <marex@denx.de>
mx6solox has a requirement for 64 bytes alignment for RX DMA transfer.
Other SoCs work with the standard 32 bytes alignment.
Adjust it accordingly by using 64 bytes aligment in the FEC RX DMA buffers,
which addresses the needs from mx6solox and also works for the other SoCs.
Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Acked-by: Marek Vasut <marex@denx.de>
This patch is to add etsec support for LS102xA. First, Little-endian
descriptor mode should be enabled. So RxBDs and TxBDs are interpreted
with little-endian byte ordering. Second, TSEC_SIZE and TSEC_MDIO_OFFSET
are different from PowerPC, redefine them for LS1021xA.
Signed-off-by: Alison Wang <alison.wang@freescale.com>
fsl_enet.h defines the mapping of the usual MII management
registers, which are included in the MDIO register block
common to Freescale ethernet controllers. So it shouldn't
depend on the CPU architecture but it should be actually
part of the arch independent fsl_mdio.h.
To remove the arch dependency, merge the content of
asm/fsl_enet.h into fsl_mdio.h.
Some files (like fm_eth.h) were simply including fsl_enet.h
only for phy.h. These were updated to include phy.h instead.
Signed-off-by: Claudiu Manoil <claudiu.manoil@freescale.com>
There is no reason to redefine pure readl/writel functions.
So remove this redundancy.
Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@ti.com>
Acked-by: Vitaly Andrianov <vitalya@ti.com>
The Broadcom StarFighter2 Ethernet driver is used in multiple Broadcom
SoC(s) and:
- supports multiple MAC blocks,
- provides support for the Broadcom GMAC.
This driver requires MII and PHYLIB.
Signed-off-by: Jiandong Zheng <jdzheng@broadcom.com>
Signed-off-by: Steve Rae <srae@broadcom.com>
We noticed on the DXR2 platform (AM335x with a SMSC LAN9303 switch connected
to the CPSW MAC) that the network performance in U-Boot is quite poor. Only
when the transfer is started without a cable connected, and the cable is
plugged after the first timeout "T" occured, an increased in performance
can be seen. Debugging has revealed, that the cpsw driver has constant
link checking builtin into the rx and tx functions. This results in the
bad performance and seems to be unnecessary. The link has already been
checked in the init function, before the transfer is started. This usually
is sufficient.
BTW: I have seen no other network driver in U-Boot so far, that constantly
checks for link in the rx / tx functions.
The performance numbers on the DXR2 board are:
0.56 MiB/s cpsw_check_link() in rx and tx path
0.87 MiB/s cpsw_check_link() only in tx path
1.0 MiB/s cpsw_check_link() only in rx path
2.7 MiB/s no cpsw_check_link() in rx and tx path
So with this patch the network performance on DXR2 increases from 0.56
to 2.7 MiB/s (nearly 5 times as fast).
Signed-off-by: Stefan Roese <sr@denx.de>
Tested-by: Samuel Egli <samuel.egli@siemens.com>
Tested-by: Heiko Schocher <hs@denx.de>
Cc: Vladimir Koutny <vladimir.koutny@streamunlimited.com>
Cc: Mugunthan V N <mugunthanvnm@ti.com>
Cc: Joe Hershberger <joe.hershberger@gmail.com>
Cc: Tom Rini <trini@ti.com>
Implement proper support for cache flushing and invalidation into the
Intel e1000 NIC driver.
Signed-off-by: Marek Vasut <marex@denx.de>
Acked-by: Tim Harvey <tharvey@gateworks.com>
The following configs are not defined at all:
- CONFIG_INCA_IP_SWITCH
- CONFIG_PBL2800_ETHER
- CONFIG_PHY_ICPLUS
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
XFI is supported on T4QDS-XFI board, which removed slot3, and four LANEs
of serdes2 are routed to a SFP+ cages, which to house fiber cable or
direct attach cable(copper), the copper cable is used to emulate the
10GBASE-KR scenario.
So, for XFI usage, there are two scenarios, one will use fiber cable,
another will use copper cable. For fiber cable, there is NO PHY, while
for copper cable, we need to use internal PHY which exist in Serdes to
do auto-negotiation and link training, which implemented in kernel.
We use hwconfig to define cable type for XFI, and fixup dtb based on the
cable type.
For copper cable, set below env in hwconfig:
fsl_10gkr_copper:<10g_mac_name>
the <10g_mac_name> can be fm1_10g1, fm1_10g2, fm2_10g1, fm2_10g2. The
four <10g_mac_name>s do not have to be coexist in hwconfig. For XFI ports,
if a given 10G port will use the copper cable for 10GBASE-KR, set the
<10g_mac_name> of the port in hwconfig, otherwise, fiber cable will be
assumed to be used for the port.
For ex. if four XFI ports will both use copper cable, the hwconfig
should contain:
fsl_10gkr_copper:fm1_10g1,fm1_10g2,fm2_10g1,fm2_10g2
For fiber cable:
1. give PHY address to a XFI port, otherwise, the XFI ports will not be
available in U-boot, there is no PHY physically for XFI when using fiber
cable, this is just to make U-boot happy and we can use the XFI ports
in U-boot.
2. fixup dtb to use fixed-link in case of fiber cable which has no PHY.
Kernel requests that a MAC must have a PHY or fixed-link.
When using XFI protocol, the MAC 9/10 on FM1 should init as 10G interface.
Change serdes 2 protocol 56 to 55 which has same feature as 56 since 56
is not valid any longer.
Signed-off-by: Shaohui Xie <Shaohui.Xie@freescale.com>
Reviewed-by: York Sun <yorksun@freescale.com>
Messages to afleming@freescale.com now bounce, and should be
directed to my personal address at afleming@gmail.com
Signed-off-by: Andy Fleming <afleming@gmail.com>
R8A7794 has the same sh-ether IP core as other SH/rmobile.
This patch adds support of R8A7794.
Signed-off-by: Nobuhiro Iwamatsu <nobuhiro.iwamatsu.yj@renesas.com>
At present stdio device functions do not get any clue as to which stdio
device is being acted on. Some implementations go to great lengths to work
around this, such as defining a whole separate set of functions for each
possible device.
For driver model we need to associate a stdio_dev with a device. It doesn't
seem possible to continue with this work-around approach.
Instead, add a stdio_dev pointer to each of the stdio member functions.
Note: The serial drivers have the same problem, but it is not strictly
necessary to fix that to get driver model running. Also, if we convert
serial over to driver model the problem will go away.
Code size increases by 244 bytes for Thumb2 and 428 for PowerPC.
22: stdio: Pass device pointer to stdio methods
arm: (for 2/2 boards) all +244.0 bss -4.0 text +248.0
powerpc: (for 1/1 boards) all +428.0 text +428.0
Signed-off-by: Simon Glass <sjg@chromium.org>
Acked-by: Marek Vasut <marex@denx.de>
Reviewed-by: Marek Vasut <marex@denx.de>
Some of the fm_port_to_index() callers did not check for -1 return value and
used -1 as an array index.
Signed-off-by: Marian Rotariu <marian.rotariu@freescale.com>
Reviewed-by: York Sun <yorksun@freescale.com>
There have been 3 versions of the sunxi_emac support patch during its
development. Somehow version 2 ended up in upstream u-boot where as
the u-boot-sunxi git repo got version 3.
This bumps the version in upstream u-boot to version 3 of the patch:
- Initialize MII clock earlier so mii access to allow independent use
- Name change from WEMAC to EMAC to match mainline kernel & chip manual
- Cosmetic code cleanup
Signed-off-by: Stefan Roese <sr@denx.de>
Signed-off-by: Henrik Nordstrom <henrik@henriknordstrom.net>
Signed-off-by: Oliver Schinagl <oliver@schinagl.nl>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Ian Campbell <ijc@hellion.org.uk>
Adding support to load and start the Layerscape Management Complex (MC)
firmware. First, the MC GCR register is set to 0 to reset all cores. MC
firmware and DPL images are copied from their location in NOR flash to
DDR. MC registers are updated with the location of these images.
Deasserting the reset bit of MC GCR register releases core 0 to run.
Core 1 will be released by MC firmware. Stop bits are not touched for
this step. U-boot waits for MC until it boots up. In case of a failure,
device tree is updated accordingly. The MC firmware image uses FIT format.
Signed-off-by: J. German Rivera <German.Rivera@freescale.com>
Signed-off-by: York Sun <yorksun@freescale.com>
Signed-off-by: Lijun Pan <Lijun.Pan@freescale.com>
Signed-off-by: Shruti Kanetkar <Shruti@Freescale.com>
Add to code to flush the dcache after we writing in DMA buffer.
Also we need invalidate the dcache before we check the status in the
DMA buffer.
Tested in SAMA5D3x-EK with gmac0. Tftp download speed shows in below:
Disable DCache: 1.1 MiB/s
Enable DCache: 1.6 MiB/s
Increase speed with about 40%.
The code should have no impact with the boards which are not
enable_dcache().
Tested in AT91SAM9M10G45EK.
Signed-off-by: Josh Wu <josh.wu@atmel.com>
Signed-off-by: Andreas Bießmann <andreas.devel@googlemail.com>
This also renames the CONFIG_SYS_MACB_xx defines. They are used just local and
therefore don't need the CONFIG_SYS_ prefix.
Signed-off-by: Andreas Bießmann <andreas.devel@googlemail.com>
Reviewed-by: Josh Wu <josh.wu@atmel.com>
'r' of rESR_RTLF is a mistake of E.
Signed-off-by: Nobuhiro Iwamatsu <nobuhiro.iwamatsu.yj@renesas.com>
Signed-off-by: Nobuhiro Iwamatsu <iwamatsu@nigauri.org>
The R7S72100 of ARM SoC that Renesas manufactured has one Ether port.
This has the same IP SH-Ether. This patch adds support of the R7S72100
in SH-Ether.
Signed-off-by: Nobuhiro Iwamatsu <nobuhiro.iwamatsu.yj@renesas.com>
Signed-off-by: Nobuhiro Iwamatsu <iwamatsu@nigauri.org>
get_phy_id returns -EIO when it can't read from a phy at a given addr. This would cause
create_phy_by_mask to return prematurely before it had tested the other addresses in the provided mask.
Example usage:
Replace
phydev = phy_connect(bus, phy_addr, dev, phy_if)
with
phydev = phy_find_by_mask(bus, phy_mask, phy_if)
if (phydev)
phy_connect_dev(phydev, dev);
Signed-off-by: Cormier, Jonathan <jcormier@criticallink.com>
Cc: Joe Hershberger <joe.hershberger@gmail.com>
The correct value for this setting can vary across SoCs and boards, so make it
configurable.
Also reduce the default value to 8, which is the same default as used in the
Linux driver.
Signed-off-by: Ian Campbell <ijc@hellion.org.uk>
Cc: Alexey Brodkin <abrodkin@synopsys.com>
On Thu, 2014-05-08 at 22:26 +0100, Ian Campbell wrote:
> The {r,t}xbuffs fields also need to be aligned. Previously this was done
> implicitly because they immediately followed the descriptor tables. Make this
> explicit and also move to the head of the struct.
Looks like I managed to not actually commit the move of the field to the
head of the struct! v3.1 follows....
Ian.
8<------------
>From 2937ba01841887317f6792709ed57cb86b5fc0cd Mon Sep 17 00:00:00 2001
From: Ian Campbell <ijc@hellion.org.uk>
Date: Thu, 1 May 2014 19:45:15 +0100
Subject: [PATCH] net/designware: reorder struct dw_eth_dev to pack more
efficiently.
The {tx,rx}_mac_descrtable fields are aligned to ARCH_DMA_MINALIGN, which could
be 256 or even larger. That means there is a potentially huge hole in the
struct before those fields, so move them to the front where they are better
packed.
Moving them to the front also helps ensure that so long as dw_eth_dev is
properly aligned (which it is since "net/designware: ensure device private data
is DMA aligned.") the {tx,rx}_mac_descrtable will be too, or at least avoids
having to worry too much about compiler specifics.
The {r,t}xbuffs fields also need to be aligned. Previously this was done
implicitly because they immediately followed the descriptor tables. Make this
explicit and also move to the head of the struct.
Signed-off-by: Ian Campbell <ijc@hellion.org.uk>
Cc: Alexey Brodkin <abrodkin@synopsys.com>
Tested-by: Siarhei Siamashka <siarhei.siamashka@gmail.com>
Reviewed-by: Siarhei Siamashka <siarhei.siamashka@gmail.com>
This is required at least on ARM.
When sending instead of simply invalidating the entire descriptor, flush
as little as possible while still respecting ARCH_DMA_MINALIGN, as
requested by Alexey.
Signed-off-by: Ian Campbell <ijc@hellion.org.uk>
Cc: Alexey Brodkin <abrodkin@synopsys.com>
struct dw_eth_dev contains fields which are accessed via DMA, so make sure it
is aligned to a dma boundary. Without this I see:
ERROR: v7_dcache_inval_range - start address is not aligned - 0x7fb677e0
Signed-off-by: Ian Campbell <ijc@hellion.org.uk>
Reviewed-by: Alexey Brodkin <abrodkin@synopsys.com>
Acked-by: Marek Vasut <marex@denx.de>
This patch adds support for VSC8664 PHY module which can
be found on Freescale's T4240RDB boards.
Signed-off-by: Chunhe Lan <Chunhe.Lan@freescale.com>
Reviewed-by: York Sun <yorksun@freescale.com>
The T4080 SoC is a low-power version of the T4160.
T4080 combines 4 dual-threaded Power Architecture e6500
cores with single cluster and two memory complexes.
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
Current driver uses a Maximum value for MDIO_HOLD when doing 10G MDIO
access, this is due to an errata A-006260 on T4 rev1.0 which is fixed
on rev2.0, so remove the maximum value to use the default value for rev2.0.
Signed-off-by: Shaohui Xie <Shaohui.Xie@freescale.com>
This sets up the linkage from the phydev back to the ethernet device. This
symptom of not doing this which I noticed was:
<NULL> Waiting for PHY auto negotiation to complete....
rather than:
dwmac.1c50000 Waiting for PHY auto negotiation to complete....
Signed-off-by: Ian Campbell <ijc@hellion.org.uk>
Cc: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
Some functions in include/net.h are ported from
include/linux/etherdevice.h of Linux Kernel.
For ex.
is_zero_ether_addr()
is_multicast_ether_addr()
is_broadcast_ether_addr()
is_valid_ether_addr();
So, we should use the same function name as that of Linux Kernel,
eth_rand_addr(), for consistency.
Besides, eth_rand_addr() has been implemented as an inline function.
So it should not be surrounded by #ifdef CONFIG_RANDOM_MACADDR.
Signed-off-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
As some PHYs have non-standard PHY ID registers, PHY Id can't
be read correctly by current get_phy_id function, so we enable
get_phy_id redefinable to permit specific PHY driver having own
specific get_phy_id function.
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
Function "genphy_parse_link()" used "if (mii_reg & BMSR_ANEGCAPABLE)" before,
but used "if (phydev->supported & SUPPORTED_Autoneg)" now.
So assign "phydev->supported" to "phydev->drv->features" for ar8035
to enable autonegotiation. Then removed the genphy_config_aneg() function.
Signed-off-by: Xie Xiaobo <X.Xie@freescale.com>
Add missing header.
Warnings:
drivers/net/zynq_gem.c:491:5: warning: symbol 'zynq_gem_initialize' was not declared. Should it be static?
drivers/net/zynq_gem.c:542:5: warning: symbol 'zynq_gem_of_init' was not declared. Should it be static?
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
