As explained in commit 3bedbcc3aa ("arm: mvebu: a38x: serdes: Don't
overwrite read-only SAR PCIe registers") it is required to set Maximum Link
Width bits of PCIe Root Port Link Capabilities Register depending of number
of used serdes lanes. As this register is part of PCIe address space and
not serdes address space, move it into pci_mvebu.c driver.
Read number of PCIe lanes from DT property "num-lanes" which is used also
by other PCIe controller drivers in Linux kernel. If this property is
absent then it defaults to 1. This property needs to be set to 4 for every
mvebu board which use PEX_ROOT_COMPLEX_X4 or PEX_BUS_MODE_X4.
Enabling of PCIe port needs to be done afer all registers in PCIe address
space are properly configure. For this purpose use new mvebu-reset driver
(part of system-controller) and remove this code from serdes code.
Because some PCIe ports cannot be enabled individually, it is required to
first setup all PCIe ports and then enable them.
This change contains also all required "num-lanes" and "resets" DTS
properties, to make pci_mvebu.c driver work correctly.
Signed-off-by: Pali Rohár <pali@kernel.org>
After function mvebu_pcie_probe() returns U-Boot DM expects that PCIe link
is already up. In followup patches link initialization will be moved from
SPL to proper and therefore explicitly link up delay is required.
Delay mvebu_pcie_probe() for 100ms to ensure that PCIe link is up after
function finish. In the case when no card is connected to the PCIe slot,
this will delay probe time by 100ms, which should not be problematic.
This change fixes detection and initialization of some QCA98xx cards on
the first serdes when configured in x1 mode. Default configuration of
the first serdes on A385 is x4 mode, so it looks as if some delay is
required when x4 is changed to x1 and card correctly links with A385.
Other PCIe serdes ports on A385 are x1-only, and so they don't have this
problem.
Signed-off-by: Pali Rohár <pali@kernel.org>
In first phase just parse DT properties and fill struct mvebu_pcie. In
second phase setup all PCIe links (without enabling them). And in the last
third phase enable all PCIe links and create UCLASS_PCI device for each
one.
Because parsing of DT is done before UCLASS_PCI is created, we cannot use
DM for this action anymore. So remove .of_to_plat callback and replace it
by ad-hoc function for parsing DT properties and filling struct mvebu_pcie.
Signed-off-by: Pali Rohár <pali@kernel.org>
SoC specific macro SOC_REGS_PHY_BASE is used for two things:
* calculation of base PCIe port address
* filling PCIe register with address of internal registers
For calculating base PCIe port address use function
ofnode_translate_address() which translates DT "assigned-addresses" to
final PCIe port address.
And for calculating address of internal registers use untranslated and
translated DT "assigned-addresses".
Basically this change reads SOC_REGS_PHY_BASE address indirectly from DT.
Signed-off-by: Pali Rohár <pali@kernel.org>
Function mvebu_pcie_port_parse_dt() is called only from
mvebu_pcie_of_to_plat() function. Both these function parse DT properties
required to setup mvebu pcie. So inline mvebu_pcie_port_parse_dt() function
into mvebu_pcie_of_to_plat() to have all code related to parsing DT
properties at one place.
Signed-off-by: Pali Rohár <pali@kernel.org>
Do not call pci_set_region() for resources which were not properly mapped.
This prevents U-Boot to access unmapped memory space.
Update MBUS_PCI_MEM_SIZE and MBUS_PCI_IO_SIZE macros to cover all PCIe MEM
and IO ranges. Previously these macros covered only address ranges for the
first PCIe port. Between MBUS_PCI_IO_BASE and MBUS_PCI_MEM_BASE there is
space for six 128 MB long address ranges. So set MBUS_PCI_MEM_SIZE to value
of 6*128 MB. Similarly set MBUS_PCI_IO_SIZE to 6*64 KB.
Function resource_size() returns zero when start address is 0 and end
address is -1. So set invalid resources to these values to indicate that
resource has no mapping.
Split global PCIe MEM and IO resources (defined by MBUS_PCI_*_* macros)
into PCIe ports in mvebu_pcie_bind() function which allocates per-port
based struct mvebu_pcie, instead of using global state variables
mvebu_pcie_membase and mvebu_pcie_iobase. This makes pci_mvebu.c driver
independent of global static variables (which store the state of
allocation) and allows to bind and unbind the driver more times.
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>
PCI mvebu driver uses extended format of Config Address for PCI
Configuration Mechanism #1.
So use new U-Boot macro PCI_CONF1_EXT_ADDRESS() and remove old custom
driver address macros.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Simon Glass <sjg@chromium.org>
The global data pointer is not used in this driver, remove it's
declaration.
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Use more appropriate resource_size() function when working with data in
struct resource.
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Function mvebu_pcie_setup_wins() sets up all other BARs, so move setup of
BAR[0] to this function to have common code at one place.
In the past, commit 193a1e9f19 ("pci: pci_mvebu: set BAR0 after memory
space is set") moved setup of BAR[0] to another location, due to ath10k
not working in kernel, but the reason why was unknown, but it seems to
work now, and we think the issue then was cause by the PCIe Root Port
presenting itself as a Memory Controller and therefore U-Boot's code
have overwritten the BAR. Since the driver now ignores any write
operations to PCIe Root Port BARs, this should not be an issue anymore.
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
There is no maintainer entry for pci-aardvark.c. Add entry for
pci-aardvark.c and pci_mvebu.c with Pali and Stefan as maintainers.
Signed-off-by: Pali Rohár <pali@kernel.org>
Acked-by: Stefan Roese <sr@denx.de>
Reviewed-by: Stefan Roese <sr@denx.de>
The PCI Bridge which represents mvebu PCIe Root Port has Expansion ROM
Base Address register at offset 0x30 but its meaning is different that
of PCI's Expansion ROM BAR register, although the address format of
the register is the same.
In reality, this device does not have any configurable PCI BARs. So
ensure that write operation into BARs (including Expansion ROM BAR) is a
noop and registers always contain zero address which indicates that BARs
are unsupported.
Fixes: a7b61ab58d ("pci: pci_mvebu: Properly configure and use PCI Bridge (PCIe Root Port)")
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Since u32 takes up 4 bytes, we need to divide the number of u32s by 4
for cfgcache.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Header file mach/cpu.h already defines MBUS_PCI_MEM_SIZE macro which
defines size of MBUS_PCI_MEM_BASE window. So use global MBUS_PCI_MEM_SIZE
macro instead of locally defined PCIE_MEM_SIZE macro. Both macros have same
definition.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
This is a pci driver, not an eth driver.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Root Complex should be the default mode, let's set it explicitly.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Now that PCI Bridge is working, U-Boot's CONFIG_PCI_PNP code automatically
enables memory access and bus mastering when it is needed. So do not
prematurely enable memory access and bus mastering.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
PCI Bridge is always accessible also when link is down. So move detection
of link up from mvebu_pcie_of_to_plat() function to mvebu_pcie_valid_addr()
function which is used when accessing PCI config space.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Functions mvebu_pcie_get_local_bus_nr() and mvebu_pcie_get_local_dev_nr()
are not used, so remove them.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
The mysterious "Memory controller" PCI device which is present in PCI
config space is improperly configured and crippled PCI Bridge which acts
as PCIe Root Port for endpoint PCIe card.
This PCI Bridge reports in PCI config space incorrect Class Code (Memory
Controller) and incorrect Header Type (Type 0). It looks like HW bug in
mvebu PCIe controller but apparently it can be changed via mvebu registers
to correct values.
The worst thing is that this PCI Bridge is crippled and its PCI config
registers in range 0x10-0x34 alias access to internal mvebu registers which
have different functionality as PCI Bridge registers. Moreover,
configuration of PCI primary and secondary bus numbers (registers 0x18
and 0x19) is done via totally different mvebu registers via totally strange
method and cannot be done via PCI Bridge config space.
Due to above fact about PCI config range 0x10-0x34, allocate a private
cfgcache[] buffer in the driver, to which PCI config access requests to
the 0x10-0x34 space will be redirected in mvebu_pcie_read_config() and
mvebu_pcie_write_config() functions. Function mvebu_pcie_write_config()
will also catch writes to PCI_PRIMARY_BUS (0x18) and PCI_SECONDARY_BUS
(0x19) registers and set PCI Bridge primary and secondary bus numbers via
mvebu's own method.
Also, Expansion ROM Base Address register (0x38) is available, but at
different offset 0x30. So recalculate register offset before accessing PCI
config space.
After these steps U-Boot sees working PCI Bridge and CONFIG_PCI_PNP code
can finally start enumerating all PCIe devices correctly, even with more
complicated PCI topology. So update also mvebu_pcie_valid_addr() function
to reflect state of the real device topology.
Each PCIe port is de-facto isolated and every PCI Bridge which is part of
PCIe Root Complex is also isolated, so put them on separate PCI buses as
(local) device 0.
U-Boot already supports enumerating separate PCI buses, real (HW) bus
number can be retrieved by "PCI_BUS(bdf) - dev_seq(bus)" code, so update
config read/write functions to properly handle more complicated tree
topologies (e.g. when a PCIe switch with multiple PCI buses is connected
to the PCIe port).
Local bus number and local device number on mvebu are used for determining
which config request type is used (Type 0 vs Type 1). On normal non-broken
PCIe hardware it is done by primary and secondary bus numbers. So correctly
translate settings between these numbers to ensure that correct config
requests are sent over the PCIe bus.
As bus numbers are correctly re-configured, it does not make sense to print
some initial bogus configuration during probe, so remove this debug code.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
When reading 8 or 16 bits from config space, use appropriate readb() or
readw() calls. This ensures that PCIe controller does not read more bits
from endpoint card as asked by read_config() function.
Technically there should not be an issue with reading data from config
space which are not later used as there are no clear-by-read registers.
But it is better to use correct read operation based on requested size.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Current implementation of write_config() is broken for PCI_SIZE_8 or
PCI_SIZE_16 as it always uses writel(), which means that write operation
is always 32-bit, so upper 24 bits for PCI_SIZE_8 and upper 16 bits for
PCI_SIZE_16 are cleared.
Fix this by using writeb() and writew(), respectively.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Linux displays the real PCIe card connected to a mvebu PCIe slot as
device 0, not 1. This is done by setting local dev number to 1, so that
the local "Marvell Memory controller" device is on address 1.
Let's do it also in U-Boot.
With this commit the pci command in U-Boot prints something like:
=> pci
Scanning PCI devices on bus 0
BusDevFun VendorId DeviceId Device Class Sub-Class
_____________________________________________________________
00.00.00 0x168c 0x003c Network controller 0x80
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Cc: Stefan Roese <sr@denx.de>
Cc: Phil Sutter <phil@nwl.cc>
Cc: Mario Six <mario.six@gdsys.cc>
Cc: Baruch Siach <baruch@tkos.co.il>
Reviewed-by: Stefan Roese <sr@denx.de>
Refactor validation of bdf parameter in mvebu_pcie_read/write_config
functions.
We can simplify the code by putting the validation into separate
function.
Also there are always only two devices visible on local bus:
* on slot configured by function mvebu_pcie_set_local_dev_nr()
(by default this register is set to 0) there is a
"Marvell Memory controller", which isn't useful in root complex
mode,
* on all other slots the real PCIe card connected to the PCIe slot.
We can simplify the code even more by simply allowing access only to
the real PCIe card.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Cc: Stefan Roese <sr@denx.de>
Cc: Phil Sutter <phil@nwl.cc>
Cc: Mario Six <mario.six@gdsys.cc>
Cc: Baruch Siach <baruch@tkos.co.il>
Reviewed-by: Stefan Roese <sr@denx.de>
Other drivers (aardvark, intel_fpga) print "(addr,size,val)" when
debugging is enabled. Print size for pci_mvebu as well.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Cc: Stefan Roese <sr@denx.de>
Cc: Phil Sutter <phil@nwl.cc>
Cc: Mario Six <mario.six@gdsys.cc>
Cc: Baruch Siach <baruch@tkos.co.il>
Reviewed-by: Stefan Roese <sr@denx.de>
Write bdf address in a same way in mvebu_pcie_read/write_config.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Cc: Stefan Roese <sr@denx.de>
Cc: Phil Sutter <phil@nwl.cc>
Cc: Mario Six <mario.six@gdsys.cc>
Cc: Baruch Siach <baruch@tkos.co.il>
Reviewed-by: Stefan Roese <sr@denx.de>
PCI uclass maps PCI bus numbers to the seq member of struct udevice.
Use dev_seq(dev) as the bus number in mvebu_pcie_probe instead of an
incrementing a static variable.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Cc: Stefan Roese <sr@denx.de>
Cc: Phil Sutter <phil@nwl.cc>
Cc: Mario Six <mario.six@gdsys.cc>
Cc: Baruch Siach <baruch@tkos.co.il>
Reviewed-by: Stefan Roese <sr@denx.de>
This patch changes the PCI config routines in the Armada XP / 38x driver
to not allow access to the PCIe root ports.
While updating the Armada XP based theadorable to the latest mainline
and testing it with the DM PCI driver I noticed, that the PCI root
bridge was being configured incorrectly. Resulting in the PCIe Intel
WiFi was not working correctly in Linux. With this patch applied, all
PCIe devices work without any issues in Linux again.
Signed-off-by: Stefan Roese <sr@denx.de>
Cc: Marek Behún <marek.behun@nic.cz>
Cc: Phil Sutter <phil@nwl.cc>
Cc: Mario Six <mario.six@gdsys.cc>
Move this out of the common header and include it only where needed. In
a number of cases this requires adding "struct udevice;" to avoid adding
another large header or in other cases replacing / adding missing header
files that had been pulled in, very indirectly. Finally, we have a few
cases where we did not need to include <asm/global_data.h> at all, so
remove that include.
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Rini <trini@konsulko.com>
Configure an IO region and window for PNP identical to how MEM region is
set up. Linux does this only if the DT defines a pcie-io-aperture
property for the SOC, but since all supported boards do this should not
be needed.
Signed-off-by: Phil Sutter <phil@nwl.cc>
Reviewed-by: Stefan Roese <sr@denx.de>
This name is far too long. Rename it to remove the 'data' bits. This makes
it consistent with the platdata->plat rename.
Signed-off-by: Simon Glass <sjg@chromium.org>
We use 'priv' for private data but often use 'platdata' for platform data.
We can't really use 'pdata' since that is ambiguous (it could mean private
or platform data).
Rename some of the latter variables to end with 'plat' for consistency.
Signed-off-by: Simon Glass <sjg@chromium.org>
This construct is quite long-winded. In earlier days it made some sense
since auto-allocation was a strange concept. But with driver model now
used pretty universally, we can shorten this to 'auto'. This reduces
verbosity and makes it easier to read.
Coincidentally it also ensures that every declaration is on one line,
thus making dtoc's job easier.
Signed-off-by: Simon Glass <sjg@chromium.org>
At present dm/device.h includes the linux-compatible features. This
requires including linux/compat.h which in turn includes a lot of headers.
One of these is malloc.h which we thus end up including in every file in
U-Boot. Apart from the inefficiency of this, it is problematic for sandbox
which needs to use the system malloc() in some files.
Move the compatibility features into a separate header file.
Signed-off-by: Simon Glass <sjg@chromium.org>
At present this method uses a non-const udevice pointer, but the call
should not modify the device. Use a const pointer.
Signed-off-by: Simon Glass <sjg@chromium.org>
The non-DM version of this driver used to set BAR0 register after the
calls to pci_set_region.
I found out that for some strange reason the ath10k driver in kernel
fails to work if this is done the other way around.
I know that Linux's driver should not depend on how U-Boot does things,
but for some strange reason it does and this seems to be the simplest
solution. Fix it since it caused regressions on Omnia.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Cc: Stefan Roese <sr@denx.de>
Cc: Dirk Eibach <dirk.eibach@gdsys.cc>
Cc: Mario Six <mario.six@gdsys.cc>
Cc: Chris Packham <chris.packham@alliedtelesis.co.nz>
Cc: Phil Sutter <phil@nwl.cc>
Cc: VlaoMao <vlaomao@gmail.com>
Tested-by: Chris Packham <judge.packham@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Stefan Roese <sr@denx.de>
This patch adds a comment to explain the use of the hardcoded value for
the number of address cells in mvebu_get_tgt_attr(). This should help to
rework this function, once CONFIG_OF_LIVE is enabled for MVEBU in
general.
Signed-off-by: Stefan Roese <sr@denx.de>
Cc: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Stefan Roese <sr@denx.de>
This patch adds DM_PCI support to the MVEBU PCIe driver. This is
necessary, since all PCI drivers have to be moved to DM (driver model)
until the v2019.07 release.
To not break git bisect'ablility, this patch also moves CONFIG_PCI_MVEBU
from config headers to the defconfig files.
Signed-off-by: Stefan Roese <sr@denx.de>
Cc: Dirk Eibach <dirk.eibach@gdsys.cc>
Cc: Mario Six <mario.six@gdsys.cc>
Cc: Chris Packham <chris.packham@alliedtelesis.co.nz>
Cc: Phil Sutter <phil@nwl.cc>
Cc: Marek Behún <marek.behun@nic.cz>
Cc: VlaoMao <vlaomao@gmail.com>
When U-Boot started using SPDX tags we were among the early adopters and
there weren't a lot of other examples to borrow from. So we picked the
area of the file that usually had a full license text and replaced it
with an appropriate SPDX-License-Identifier: entry. Since then, the
Linux Kernel has adopted SPDX tags and they place it as the very first
line in a file (except where shebangs are used, then it's second line)
and with slightly different comment styles than us.
In part due to community overlap, in part due to better tag visibility
and in part for other minor reasons, switch over to that style.
This commit changes all instances where we have a single declared
license in the tag as both the before and after are identical in tag
contents. There's also a few places where I found we did not have a tag
and have introduced one.
Signed-off-by: Tom Rini <trini@konsulko.com>
Armada 38x has four PCI ports, not three.
The optimization in pci_init_board() seems to assume that every port has
three lanes. This is obviously wrong, and breaks support for Armada 38x.
Signed-off-by: Dirk Eibach <dirk.eibach@gdsys.cc>
Signed-off-by: Mario Six <mario.six@gdsys.cc>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Stefan Roese <sr@denx.de>
Now, arch/${ARCH}/include/asm/errno.h and include/linux/errno.h have
the same content. (both just wrap <asm-generic/errno.h>)
Replace all include directives for <asm/errno.h> with <linux/errno.h>.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
[trini: Fixup include/clk.]
Signed-off-by: Tom Rini <trini@konsulko.com>
Armada XP has support for X4 lanes, boards specify this in their
serdes_cfg. During PEX init in high_speed_env_lib.c, the configuration
is stored in GEN_PURP_RES_2_REG.
When enumerating PEX, subsequent interfaces of an X4 lane must be
skipped. Otherwise the enumeration hangs up the board.
The way this is implemented here is not exactly beautiful, but it mimics
how Marvell's BSP does it. Alternatively we could get the information
using board_serdes_cfg_get(), but that won't lead to clean code, either.
Signed-off-by: Phil Sutter <phil@nwl.cc>
Acked-by: Stefan Roese <sr@denx.de>
Reviewed-by: Tom Rini <trini@konsulko.com>
This adds a PCI driver for the controllers found on Marvell MVEBU SoCs.
Besides the driver, this patch also removes the statically defined
PCI MBUS windows. As they are not needed anymore, since this PCIe
driver now creates the windows dynamically.
Tested on Armada XP db-mv784mp-gp eval board using an Intel E1000
PCIe card in all 3 PCIe slots. And on the Armada 38x db-88f6820-gp
eval board using this Intel E1000 PCIe card in the PCIe 0 slot.
This port was done in cooperation with Anton Schubert.
Signed-off-by: Anton Schubert <anton.schubert@gmx.de>
Signed-off-by: Stefan Roese <sr@denx.de>
Cc: Luka Perkov <luka.perkov@sartura.hr>
Cc: Dirk Eibach <eibach@gdsys.de>
