pandora-u-boot.git
2 years agoarm: a37xx: pci: Fix configuring PCIe resources
Pali Rohár [Wed, 26 May 2021 15:59:40 +0000 (17:59 +0200)]
arm: a37xx: pci: Fix configuring PCIe resources

The `ranges` DT property of the PCIe node is currently ignored by
Aardvark driver - all entries are used as transparent PCIe MEM, despite
some of them being defined for IO in DT.

This is because the driver does not setup PCIe outbound windows and thus
a default configuration is used.

This can cause an external abort on CPU when a device driver tries to
access non-MEM space.

Setup the PCIe windows according to the `ranges` property for all
non-MEM resources (currently only IO) and also non-transparent MEM
resources.

Because Linux expects that bootloader does not setup Aardvark PCIe
windows, disable them before booting Linux.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoarm: a37xx: pci: Increase PCIe MEM size from 16 MiB to 127 MiB
Pali Rohár [Wed, 26 May 2021 15:59:39 +0000 (17:59 +0200)]
arm: a37xx: pci: Increase PCIe MEM size from 16 MiB to 127 MiB

For some configurations with more PCIe cards and PCIe bridges, 16 MiB of
PCIe MEM space may not be enough. Since TF-A already allocates a 128 MiB
CPU window for PCIe, and since IO port space is only 64 KiB in total,
use all the remaining space (64 + 32 + 16 + 8 + 4 + 2 + 1 = 127 MiB) for
PCIe MEM.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoarm: a37xx: pci: Fix a3700_fdt_fix_pcie_regions() function
Pali Rohár [Wed, 26 May 2021 15:59:38 +0000 (17:59 +0200)]
arm: a37xx: pci: Fix a3700_fdt_fix_pcie_regions() function

Current version of this function uses a lot of incorrect assumptions about
the `ranges` DT property:

 * parent(#address-cells) == 2
 * #size-cells == 2
 * number of entries == 2
 * address size of first entry == 0x1000000
 * second child address entry == base + 0x1000000

Trying to increase PCIe MEM space to more than 16 MiB leads to an overlap
with PCIe IO space, and trying to define additional MEM space (as a third
entry in the `ranges` DT property) causes U-Boot to crash when booting the
kernel.

  ## Flattened Device Tree blob at 04f00000
     Booting using the fdt blob at 0x4f00000
     Loading Device Tree to 000000001fb01000, end 000000001fb08f12 ... OK
  ERROR: board-specific fdt fixup failed: <unknown error>
   - must RESET the board to recover.

Fix a3700_fdt_fix_pcie_regions() to properly parse and update all addresses
in the `ranges` property according to
https://elinux.org/Device_Tree_Usage#PCI_Address_Translation

Now it is possible to increase PCIe MEM space from 16 MiB to maximal value
of 127 MiB.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Fixes: cb2ddb291ee6 ("arm64: mvebu: a37xx: add device-tree fixer for PCIe regions")
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoarm: a37xx: pci: Find PCIe controller node by compatible instead of path
Pali Rohár [Wed, 26 May 2021 15:59:37 +0000 (17:59 +0200)]
arm: a37xx: pci: Find PCIe controller node by compatible instead of path

Find PCIe DT node by compatible string instead of retrieving it by using
hardcoded DT path.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoarm: a37xx: pci: Fix DT compatible string to Linux' DT compatible
Pali Rohár [Wed, 26 May 2021 15:59:36 +0000 (17:59 +0200)]
arm: a37xx: pci: Fix DT compatible string to Linux' DT compatible

Change DT compatible string for A3700 PCIe from 'marvell,armada-37xx-pcie'
to 'marvell,armada-3700-pcie' to make U-Boot A3700 PCIe DT node compatible
with Linux' DT node.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoarm: a37xx: pci: Disable bus mastering when unloading driver
Pali Rohár [Wed, 26 May 2021 15:59:35 +0000 (17:59 +0200)]
arm: a37xx: pci: Disable bus mastering when unloading driver

Disable Root Bridge I/O space, memory space and bus mastering in Aardvark's
remove method, which is called before booting Linux kernel.

This ensures that PCIe device which was initialized and used by U-Boot
cannot do new DMA transfers until Linux initializes PCI subsystem and loads
appropriate drivers for the device.

During initialization of PCI subsystem Linux in fact disables this bus
mastering on Root Bridge (and later enables it when driver is loaded and
configured), but there is a possibility of a small window after U-Boot
boots Linux when bus mastering is enabled, which is not correct.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoarm: a37xx: pci: Don't put link into LTSSM Recovery state during probe
Pali Rohár [Wed, 26 May 2021 15:59:34 +0000 (17:59 +0200)]
arm: a37xx: pci: Don't put link into LTSSM Recovery state during probe

During our debugging of the Aardvark driver in Linux we have discovered
that the PCIE_CORE_LINK_CTRL_STAT_REG register in fact controls standard
PCIe Link Control Register for PCIe Root Bridge. This led us to discover
that the name of the PCIE_CORE_LINK_TRAINING macro and the corresponding
comment by this macro's usage is misleading; this bit in fact controls
Retrain Link, which, according to PCIe base spec is defined as:

  A write of 1b to this bit initiates Link retraining by directing the
  Physical Layer LTSSM to the Recovery state. If the LTSSM is already in
  Recovery or Configuration, re-entering Recovery is permitted but not
  required.

Entering Recovery state is normally done from LTSSM L0, L0s and L1 states.
But since the pci-aardvark.c driver enables Link Training just a few lines
above, the controller is not in L0 ready state yet. So setting aardvark bit
PCIE_CORE_LINK_TRAINING does not actually enter Recovery state at this
place.

Moreover, trying to enter LTSSM Recovery state without other configuration
is causing issues for some cards (e.g. Atheros AR9xxx and QCA9xxx). Since
Recovery state is not entered, these issues are not triggered.

Remove code which tries to enter LTSSM Recovery state completely.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agofastboot: Fix overflow when calculating chunk size
Sean Anderson [Thu, 27 May 2021 16:02:34 +0000 (12:02 -0400)]
fastboot: Fix overflow when calculating chunk size

If a chunk was larger than 4GiB, then chunk_data_sz would overflow and
blkcnt would not be calculated correctly. Upgrade it to a u64 and cast
its multiplicands as well. Also fix bytes_written while we're at it.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
2 years agoMerge https://source.denx.de/u-boot/custodians/u-boot-riscv
Tom Rini [Mon, 31 May 2021 14:19:14 +0000 (10:19 -0400)]
Merge https://source.denx.de/u-boot/custodians/u-boot-riscv

- SiFive FU740 and Unmatched support

2 years agoMerge https://source.denx.de/u-boot/custodians/u-boot-sunxi
Tom Rini [Mon, 31 May 2021 14:18:26 +0000 (10:18 -0400)]
Merge https://source.denx.de/u-boot/custodians/u-boot-sunxi

This contains the fix to bring back the SD card as MMC0. In the long run
we are looking into a more robust solution, but for now we need to fix
this, as this breaks the user experience left, right, and centre.
Also add the one MAINTAINERS path addition from Samuel.

2 years agodrivers: pci: pcie_dw_common: fix Werror compilation error
Green Wan [Wed, 19 May 2021 11:16:15 +0000 (04:16 -0700)]
drivers: pci: pcie_dw_common: fix Werror compilation error

Fix compilation error when Werror is turned on. The warning could
possible break some CI builds.

Signed-off-by: Green Wan <green.wan@sifive.com>
Reviewed-by: Leo Yu-Chi Liang <ycliang@andestech.com>
2 years agoriscv: cpu: fu740: clear feature disable CSR
Green Wan [Thu, 27 May 2021 13:52:14 +0000 (06:52 -0700)]
riscv: cpu: fu740: clear feature disable CSR

Clear feature disable CSR to turn on all features of hart. The detail
is specified at section, 'SiFive Feature Disable CSR', in user manual

https://sifive.cdn.prismic.io/sifive/aee0dd4c-d156-496e-a6c4-db0cf54bbe68_sifive_U74MC_rtl_full_20G1.03.00_manual.pdf

Signed-off-by: Green Wan <green.wan@sifive.com>
Reviewed-by: Sean Anderson <seanga2@gmail.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Rick Chen <rick@andestech.com>
2 years agoboard: sifive: add HiFive Unmatched board support
Green Wan [Thu, 27 May 2021 13:52:13 +0000 (06:52 -0700)]
board: sifive: add HiFive Unmatched board support

Add defconfig and board support for HiFive Unmatched.

Signed-off-by: Green Wan <green.wan@sifive.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Rick Chen <rick@andestech.com>
2 years agoriscv: dts: add SiFive Unmatched board support
Green Wan [Thu, 27 May 2021 13:52:12 +0000 (06:52 -0700)]
riscv: dts: add SiFive Unmatched board support

Add dts files for SiFive Unmatched board.

Signed-off-by: Green Wan <green.wan@sifive.com>
Reviewed-by: Rick Chen <rick@andestech.com>
2 years agoriscv: dts: add fu740 support
Green Wan [Thu, 27 May 2021 13:52:11 +0000 (06:52 -0700)]
riscv: dts: add fu740 support

Add dts support for fu740. The HiFive Unmatched support is based on
fu740 cpu and drivers in following patch set.

Signed-off-by: Green Wan <green.wan@sifive.com>
[greentime.hu: set fu740 speed to 1.2GHz]
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Rick Chen <rick@andestech.com>
2 years agodrivers: pci: add pcie support for fu740
Green Wan [Thu, 27 May 2021 13:52:10 +0000 (06:52 -0700)]
drivers: pci: add pcie support for fu740

Add pcie driver for SiFive fu740, the driver depends on
fu740 gpio, clk and reset driver to do init. Force running at Gen1
for better capatible enumeration.

Several devices are tested:
a) M.2 NVMe SSD
b) USB-to-PCI adapter
c) Ethernet adapter (E1000 compatible)

Signed-off-by: Green Wan <green.wan@sifive.com>
Reviewed-by: Neil Armstrong <narmstrong@baylibre.com>
2 years agodrivers: ram: sifive: rename fu540_ddr and add fu740 support
Green Wan [Thu, 27 May 2021 13:52:09 +0000 (06:52 -0700)]
drivers: ram: sifive: rename fu540_ddr and add fu740 support

Rename fu540_ddr.c to sifive_ddr.c and add fu740 support

Signed-off-by: Green Wan <green.wan@sifive.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
2 years agodrivers: clk: add fu740 support
Green Wan [Thu, 27 May 2021 13:52:08 +0000 (06:52 -0700)]
drivers: clk: add fu740 support

Add fu740 support. One abstract layer is added for supporting
multiple chips such as fu540 and fu740.

Signed-off-by: Green Wan <green.wan@sifive.com>
2 years agoriscv: cpu: fu740: Add support for cpu fu740
Green Wan [Thu, 27 May 2021 13:52:07 +0000 (06:52 -0700)]
riscv: cpu: fu740: Add support for cpu fu740

Add SiFive fu740 cpu to support RISC-V arch

Signed-off-by: Green Wan <green.wan@sifive.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
2 years agoMAINTAINERS: Add allwinner/sunxi driver directories
Samuel Holland [Mon, 19 Apr 2021 03:13:36 +0000 (22:13 -0500)]
MAINTAINERS: Add allwinner/sunxi driver directories

These drivers are sunxi platform-specific, and so are of interest to the
sunxi maintainers.

In fact, as there is no PHY driver maintainer, drivers/phy/allwinner had
no maintainer at all.

Signed-off-by: Samuel Holland <samuel@sholland.org>
Acked-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
2 years agosunxi: Bring back SD card as MMC device 0
Andre Przywara [Fri, 16 Apr 2021 10:52:27 +0000 (11:52 +0100)]
sunxi: Bring back SD card as MMC device 0

Commit 2243d19e5618 ("mmc: mmc-uclass: Use dev_seq() to read aliases
node's index") now actually enforces U-Boot's device enumeration policy,
where explicitly named devices come first, then any other non-named
devices follow, without filling gaps.

For quite a while we have had an "mmc1 = &mmc2;" alias in our
sunxi-u-boot.dtsi, which now leads to the problem that the SD card
(which was always mmc device 0) now gets to be number 2.
This breaks quite some boot scripts, including our own distro boot
commands, and some other features looking at $mmc_bootdev, also
fastboot.

Just add an explicit mmc0 alias in the very same file to fix this and
restore the old behaviour.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reported-by: Samuel Holland <samuel@sholland.org>
Tested-by: Simon Baatz <gmbnomis@gmail.com>
2 years agoMerge tag 'u-boot-stm32-20210528' of https://source.denx.de/u-boot/custodians/u-boot-stm
Tom Rini [Fri, 28 May 2021 18:11:06 +0000 (14:11 -0400)]
Merge tag 'u-boot-stm32-20210528' of https://source.denx.de/u-boot/custodians/u-boot-stm

- DFU: MTD: fix for lock support
- reset: stm32: fix bank bank and offset computation
- enable UNZIP config in several stm32mp defconfig

2 years agoMerge tag 'efi-2021-07-rc4-2' of https://source.denx.de/u-boot/custodians/u-boot-efi
Tom Rini [Fri, 28 May 2021 18:10:51 +0000 (14:10 -0400)]
Merge tag 'efi-2021-07-rc4-2' of https://source.denx.de/u-boot/custodians/u-boot-efi

Pull request for efi-2021-07-rc4-2

Simplify configuration using HASH functions
Fix Coverity warnings related to EFI TCG2 protocol
Enable PE/COFF image measurement

2 years agoefi_loader: add PE/COFF image measurement
Masahisa Kojima [Wed, 26 May 2021 03:09:58 +0000 (12:09 +0900)]
efi_loader: add PE/COFF image measurement

"TCG PC Client Platform Firmware Profile Specification"
requires to measure every attempt to load and execute
a OS Loader(a UEFI application) into PCR[4].
This commit adds the PE/COFF image measurement, extends PCR,
and appends measurement into Event Log.

Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Tested-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: Masahisa Kojima <masahisa.kojima@linaro.org>
Replace CONFIG_HASH_CALCULATE by CONFIG_HASH
Fix conversions between pointers and u64.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
2 years agoefi_loader: Work-around build issue due to missing hash_calculate()
Alexandru Gagniuc [Mon, 24 May 2021 19:28:57 +0000 (14:28 -0500)]
efi_loader: Work-around build issue due to missing hash_calculate()

The hash_calculate() symbol is provided by hash-checksum.c. It depends
on hash_progressive_lookup_algo(), provided when CONFIG_HASH=y.

The issue is that hash_calculate() is used by the efi_loader,
irregardless of CONFIG_FIT_SIGNATURE. As pointed out in
commit 87316da05f2f ("lib: introduce HASH_CALCULATE option"),
enabling hash_calculate() based on CONFIG_FIT_SIGNATURE is incorrect.

To resolve this, use CONFIG_HASH as the compile switch for
hash-checksum.c. This ensures that all dependencies are compiled, and
is the most natural Kconfig to use.

There is the issue of having to 'select HASH' in a couple of places
that already 'select SHA256'. This is a deeper problem with how hashes
are organized, and fixing it is beyonf the scope of this change.

Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
Acked-by: Masahisa Kojima <masahisa.kojima@linaro.org>
2 years agoRevert "lib: introduce HASH_CALCULATE option"
Alexandru Gagniuc [Mon, 24 May 2021 19:28:56 +0000 (14:28 -0500)]
Revert "lib: introduce HASH_CALCULATE option"

When we think of Kconfig, we usually think of features that we like
to enable or not. Ideally, we wouldn't use Kconfig to fix a build
issue, although sometimes it might make sense. With Kconfig it's hard
to guarantee that the fix is universal. We can only say that it works
for the set of tested configurations. In the majority of cases, it's
preferable to let the linker figure things out for us.

The reverted commit attempted to fix a build issue by adding an
invisible Kconfig option. This is wrong in several ways:

It invents a new Kconfig variable when CONFIG_HASH already
exists for the same purpose.
Second, hash-checksum.c makes use of the hash_progressive_lookup_algo()
symbol, which is only provided with CONFIG_HASH, but this dependency
was not expressed in the reverted patch.

It feels like Kconfig is turning into a listing of all available
source files, and a buffet to 'select' which ones to compile. The
purpose of this revert is to enable the next change to make use of
CONFIG_HASH instead of adding to Kconfig.

This reverts commit 87316da05f2fd49d3709275e64ef0c5980366ade.

Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
Acked-by: Masahisa Kojima <masahisa.kojima@linaro.org>
2 years agoefi_loader: Fix coverity warnings for efi tcg2 protocol
Ilias Apalodimas [Wed, 26 May 2021 18:01:00 +0000 (21:01 +0300)]
efi_loader: Fix coverity warnings for efi tcg2 protocol

Coverity reported 3 warnings on the current code.
CID 331856, 331855, 331854 on the latest scan.

Fix the rest of the warnings by initializing the variables before
passing them to tpm2_get_pcr_info().
In order to avoid future warnings and errors initialize them to 0 within
the function as well, since the values are always OR'ed after querying the
hardware.

Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
2 years agoarm: dts: stm32mp157c-odyssey-som: enable the RNG1
Grzegorz Szymaszek [Mon, 19 Apr 2021 17:55:52 +0000 (19:55 +0200)]
arm: dts: stm32mp157c-odyssey-som: enable the RNG1

Enable the true random number generator. It can be used, for example, to
generate partition UUIDs when partitioning with the gpt command. The
generator is already enabled in the device trees of several other
STM32MP1‐based boards, like DKx or DHCOM.

Signed-off-by: Grzegorz Szymaszek <gszymaszek@short.pl>
Cc: Patrice Chotard <patrice.chotard@foss.st.com>
Cc: Patrick Delaunay <patrick.delaunay@foss.st.com>
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
2 years agoconfigs: stm32mp: Enable UNZIP on STMicroelectronics stm32mp15 boards
Patrick Delaunay [Thu, 6 May 2021 07:31:00 +0000 (09:31 +0200)]
configs: stm32mp: Enable UNZIP on STMicroelectronics stm32mp15 boards

The CMD_UNZIP provides the 'gzwrite' command, which is convenient
for writing e.g. gz-compressed images to eMMC from U-Boot.

Signed-off-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
2 years agoARM: stm32: Enable UNZIP on DHSOM by default
Marek Vasut [Mon, 3 May 2021 11:31:48 +0000 (13:31 +0200)]
ARM: stm32: Enable UNZIP on DHSOM by default

The CMD_UNZIP provides the 'gzwrite' command, which is convenient
for writing e.g. gz-compressed images to eMMC from U-Boot.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Patrice Chotard <patrice.chotard@foss.st.com>
Cc: Patrick Delaunay <patrick.delaunay@foss.st.com>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
2 years agoARM: stm32: Add additional ID register check for KSZ8851 presence
Marek Vasut [Mon, 3 May 2021 11:31:39 +0000 (13:31 +0200)]
ARM: stm32: Add additional ID register check for KSZ8851 presence

Currently the code sets eth1addr only if /ethernet1 alias exists in DT,
the node pointed to by the alias has "micrel,ks8851-mll" compatible
string, and the KSZ8851 CCR register read indicates programmed EEPROM
is not connected.

This is not sufficient to detect cases where the DT still contains the
KSZ8851 nodes, but the chip itself is not present. Extend the detection
to handle these cases.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Patrice Chotard <patrice.chotard@foss.st.com>
Cc: Patrick Delaunay <patrick.delaunay@foss.st.com>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
2 years agoARM: stm32: Update dhelectronics/dh_stm32mp1/MAINTAINERS file
Christoph Niedermaier [Wed, 5 May 2021 16:23:51 +0000 (18:23 +0200)]
ARM: stm32: Update dhelectronics/dh_stm32mp1/MAINTAINERS file

Adding new DH electronics mailing list.

Signed-off-by: Christoph Niedermaier <cniedermaier@dh-electronics.com>
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
2 years agoreset: stm32: Fix bank and offset computation
Patrice Chotard [Wed, 28 Apr 2021 11:42:45 +0000 (13:42 +0200)]
reset: stm32: Fix bank and offset computation

BITS_PER_LONG is used to represent register's size which is 32.
But when compiled on arch64, BITS_PER_LONG is then equal to 64.

Fix bank and offset computation to make it work on arch32 and
arch64 and ensure that register's size is always equal to 32.

Signed-off-by: Patrice Chotard <patrice.chotard@foss.st.com>
Signed-off-by: Pankaj Dev <pankaj.dev@st.com>
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
2 years agodfu: dfu_mtd: remove the mtd_block_op error when mtd_lock is not supported
Patrick Delaunay [Wed, 10 Mar 2021 09:27:22 +0000 (10:27 +0100)]
dfu: dfu_mtd: remove the mtd_block_op error when mtd_lock is not supported

Fix the result of DFU_OP_WRITE operation in mtd_block_op function
when mtd_lock is not supported (-EOPNOTSUPP) to avoid DFU stack
error on the DFU manifestation of the MTD device, when
dfu_flush_medium_mtd is called.

Without this patch, dfu-util failed on dfuERROR state at the end
of the write operation on the alternate even if MTD write
opeartion is correctly performed.

$> dfu-util -a 3 -D test.bin
....
DFU mode device DFU version 0110
Device returned transfer size 4096
Copying data from PC to DFU device
....
Download [=========================] 100%       225469 bytes
Download done.
state(10) = dfuERROR, status(14) = Something went wrong,
  but the device does not know what it was Done!

Fixes: 65f3fc18fc1e ("dfu_mtd: Add provision to unlock mtd device")
Signed-off-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Acked-by: Sughosh Ganu <sughosh.ganu@linaro.org>
2 years agoMerge tag 'ti-v2021.07-rc4' of https://source.denx.de/u-boot/custodians/u-boot-ti
Tom Rini [Thu, 27 May 2021 11:42:49 +0000 (07:42 -0400)]
Merge tag 'ti-v2021.07-rc4' of https://source.denx.de/u-boot/custodians/u-boot-ti

- Fix reset for AM64 platforms
- Enable networking PHY driver for AM64
- Fix default R5F cluster setting in J7

2 years agoMerge branch '2021-05-26-assorted-bugfixes'
Tom Rini [Thu, 27 May 2021 11:41:25 +0000 (07:41 -0400)]
Merge branch '2021-05-26-assorted-bugfixes'

2 years agoARM: ti: Increase the allocated size for MLO.raw
Faiz Abbas [Mon, 19 Apr 2021 06:50:27 +0000 (12:20 +0530)]
ARM: ti: Increase the allocated size for MLO.raw

MLO has increased to a size greater than the allocated
128 kB in dfu_alt_info_emmc and _mmc.

Therefore, double the allocated size for MLO.raw in
the default environment.

Signed-off-by: Faiz Abbas <faiz_abbas@ti.com>
Signed-off-by: Aswath Govindraju <a-govindraju@ti.com>
Signed-off-by: Lokesh Vutla <lokeshvutla@ti.com>
Link: https://lore.kernel.org/r/20210419065027.5834-1-a-govindraju@ti.com
2 years agoarm: dts: k3-am642-sk: Add sysreset controller node
Suman Anna [Fri, 14 May 2021 01:10:57 +0000 (20:10 -0500)]
arm: dts: k3-am642-sk: Add sysreset controller node

The AM64x SoC uses a central Device Management and Security Controller
(DMSC) processor that manages all the low-level device controls
including the system-wide SoC reset. The system-wide reset is managed
through the system reset driver.

Add a sysreset controller node as a child of the dmsc node to enable
the "reset" command from U-Boot prompt for the K3 AM642 SK.

Signed-off-by: Suman Anna <s-anna@ti.com>
Signed-off-by: Lokesh Vutla <lokeshvutla@ti.com>
2 years agoarm: dts: k3-am642-evm: Add sysreset controller node
Suman Anna [Fri, 14 May 2021 01:10:56 +0000 (20:10 -0500)]
arm: dts: k3-am642-evm: Add sysreset controller node

The AM64x SoC uses a central Device Management and Security Controller
(DMSC) processor that manages all the low-level device controls
including the system-wide SoC reset. The system-wide reset is managed
through the system reset driver.

Add a sysreset controller node as a child of the dmsc node to enable
the "reset" command from U-Boot prompt for the K3 AM642 EVM.

Signed-off-by: Suman Anna <s-anna@ti.com>
Signed-off-by: Lokesh Vutla <lokeshvutla@ti.com>
2 years agofirmware: ti_sci: Update ti_sci_msg_req_reboot to include domain
Dave Gerlach [Fri, 14 May 2021 01:10:55 +0000 (20:10 -0500)]
firmware: ti_sci: Update ti_sci_msg_req_reboot to include domain

The ti_sci_msg_req_reboot message payload has been extended to include a
domain field, but for the purposes of u-boot this should be zero to
reset the entire SoC as it did before. Include domain for completeness
and set to zero to ensure proper operation.

Signed-off-by: Dave Gerlach <d-gerlach@ti.com>
Signed-off-by: Suman Anna <s-anna@ti.com>
Signed-off-by: Lokesh Vutla <lokeshvutla@ti.com>
2 years agoarm: dts: k3-j721e: Fix up MAIN R5FSS cluster mode back to Split-mode
Suman Anna [Tue, 18 May 2021 21:38:25 +0000 (16:38 -0500)]
arm: dts: k3-j721e: Fix up MAIN R5FSS cluster mode back to Split-mode

The default U-Boot environment variables and design are all set up for
both the MAIN R5FSS clusters to be in Split-mode. This is the setting
in v2021.01 U-Boot and the dt nodes are synched with the kernel binding
property names in commit 468ec2f3ef8f ("remoteproc: k3_r5: Sync to
upstreamed kernel DT property names") merged in v2021.04-rc2.

The modes for both the clusters got switched back to LockStep mode by
mistake in commit 70e167495ab2 ("arm: dts: k3-j721e: Sync Linux v5.11-rc6
dts into U-Boot") also in v2021.04-rc2. This throws the following warning
messages when early-booting the cores using default env variables,

  k3_r5f_rproc r5f at 5d00000: Invalid op: Trying to start secondary core 7 in lockstep mode
  Load Remote Processor 3 with data at addr=0x82000000 98484 bytes: Failed!
  k3_r5f_rproc r5f at 5f00000: Invalid op: Trying to start secondary core 9 in lockstep mode
  Load Remote Processor 5 with data at addr=0x82000000 98484 bytes: Failed!

Fix this by switching back both the clusters to the expected Split-mode.
Make this mode change in the u-boot specific dtsi file to avoid such
sync overrides in the future until the kernel dts is also switched to
Split-mode by default.

Fixes: 70e167495ab2 ("arm: dts: k3-j721e: Sync Linux v5.11-rc6 dts into U-Boot")
Reported-by: Minas Hambardzumyan <minas@ti.com>
Signed-off-by: Suman Anna <s-anna@ti.com>
Signed-off-by: Lokesh Vutla <lokeshvutla@ti.com>
2 years agoconfigs: am64x_evm_a53_defconfig: Enable DP83867 PHY driver
Vignesh Raghavendra [Wed, 12 May 2021 14:38:25 +0000 (20:08 +0530)]
configs: am64x_evm_a53_defconfig: Enable DP83867 PHY driver

AM64x GP and SK EVM have DP83867 PHY connected to CPSW external port0.
Enable the driver in order to use ethernet at U-Boot prompt.
CONFIG_PHY_TI is selected by CONFIG_PHY_TI_DP83867 and thus can be dropped.

Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
Signed-off-by: Lokesh Vutla <lokeshvutla@ti.com>
2 years agotest: Remove duplicate macro
Sean Anderson [Sun, 23 May 2021 17:00:56 +0000 (13:00 -0400)]
test: Remove duplicate macro

ut_asserteq_strn is defined twice. Remove one definition.

Fixes: 33d7edfd5f ("test: Add a way to check part of a console line or skip it")
Signed-off-by: Sean Anderson <seanga2@gmail.com>
Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
2 years agoGitLab: Remove tags for sandbox_noinst tests
Harm Berntsen [Fri, 21 May 2021 19:59:45 +0000 (21:59 +0200)]
GitLab: Remove tags for sandbox_noinst tests

Commit 712cc962b70d ("GitLab: Remove "tags" stanzas") removed the 'all'
tag from all the jobs. The sandbox_noinst_test.py test was added in
between the author date and commit date (commit 6c914e4232cd
("azure/gitlab: Add tests for sandbox_noinst")) which is probably why it
still contains the tags. This commit fixes that so all jobs now don't
require tags on the GitLab runners.

Signed-off-by: Harm Berntsen <harm.berntsen@nedap.com>
CC: Simon Glass <sjg@chromium.org>
Acked-by: Pratyush Yadav <p.yadav@ti.com>
2 years agotest: Fix filesystem tests always being skipped
Alper Nebi Yasak [Thu, 20 May 2021 19:09:46 +0000 (22:09 +0300)]
test: Fix filesystem tests always being skipped

Commit 1ba21bb06b08 ("test: Don't unmount not (yet) mounted system")
fixes an issue in the filesystem tests where the test setup may fail
to mount an image and still attempt to unmount it. However, the commit
unintentionally breaks the test setups in two ways.

The newly created unmounted filesystem images are being immediately
deleted due to some cleanup steps being misplaced into finally blocks,
which makes them always run instead of only on failures. The mount calls
always fail since the images never exist, causing the tests to be always
skipped. This patch moves these cleanup calls into the except blocks to
fix this and makes the tests run again.

There are also unmount calls misplaced into finally blocks, making them
run after the tests instead of before the tests. These unmount calls
make the filesystem image file consistent with the changes made to it as
part of the test setup, and this misplacement is making a number of
tests fail unexpectedly.

The unmount calls must be run before the tests use the image, meaning
before the yield call and not in the finally block. They must also be
run as a cleanup step when the filesystem setup fails, so they can't be
placed as the final call in the try blocks since they would be skipped
on such failures. For these reasons, this patch places the unmount calls
both in the except blocks and the else blocks of the final setup step.
This makes the unexpectedly failing tests to succeed again.

Furthermore, this isolates the mount calls to their own try-except
statement to avoid reintroducing the original issue of unmounting a
not-mounted image while fixing the unmount misplacement.

After these fixes, running "make tests" with guestmount available results
in two test failures not related to the mentioned commit. If the
guestmount executables are unavailable, the mounts fallback to using
sudo and result in no failures.

Fixes: 1ba21bb06b08 ("test: Don't unmount not (yet) mounted system")
Signed-off-by: Alper Nebi Yasak <alpernebiyasak@gmail.com>
Acked-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
2 years agospl: Kconfig: Use tabs instead of space for alignment
Michal Simek [Thu, 20 May 2021 11:08:44 +0000 (13:08 +0200)]
spl: Kconfig: Use tabs instead of space for alignment

A lot of entries were using spaces instead of tab for alignment that's why
it is good to fix it to use the same style everywhere.

Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agofs: btrfs: Add missing cache aligned allocation
Marek Vasut [Mon, 17 May 2021 22:39:39 +0000 (00:39 +0200)]
fs: btrfs: Add missing cache aligned allocation

The superblock buffer must be cache aligned, since it might be used
in DMA context, allocate it using ALLOC_CACHE_ALIGN_BUFFER() just
like it was done in btrfs_read_superblock() and read_tree_node().

This fixes this output on boot and non-working btrfs on iMX53:
CACHE: Misaligned operation at range [ced299d0ced2a9d0]

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Marek Behún <marek.behun@nic.cz>
Cc: Qu Wenruo <wqu@suse.com>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
2 years agofs/squashfs: zero out unused fields in fs_dirent
Heinrich Schuchardt [Mon, 17 May 2021 06:21:39 +0000 (08:21 +0200)]
fs/squashfs: zero out unused fields in fs_dirent

When reading directories the UEFI sub-system must supply file attributes
and timestamps. These fields will have to be added to struct fs_dirent.
SquashFS should not fill these fields with random data. Ensure that they
are zeroed out.

Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Reviewed-by: Miquel Raynal <miquel.raynal@bootlin.com>
2 years agopowerpc: fix regression in arch_initr_trap()
Matt Merhar [Mon, 17 May 2021 17:32:48 +0000 (17:32 +0000)]
powerpc: fix regression in arch_initr_trap()

The assembly output of the arch_initr_trap() function differed by a
single byte after common.h was removed from traps.c:

 fff49a18 <arch_initr_trap>:
 fff49a18:      94 21 ff f0     stwu    r1,-16(r1)
 fff49a1c:      7c 08 02 a6     mflr    r0
 fff49a20:      90 01 00 14     stw     r0,20(r1)
-fff49a24:      80 62 00 44     lwz     r3,68(r2)
+fff49a24:      80 62 00 38     lwz     r3,56(r2)
 fff49a28:      4b ff 76 19     bl      fff41040 <trap_init>
 fff49a2c:      80 01 00 14     lwz     r0,20(r1)
 fff49a30:      38 60 00 00     li      r3,0
 fff49a34:      38 21 00 10     addi    r1,r1,16
 fff49a38:      7c 08 03 a6     mtlr    r0

This was causing a consistent hard lockup during the MMC read / loading
of the QoriQ FMan firmware on a P2041RDB board.

Re-adding the header causes identical assembly to be emitted and allows
the firmware loading and subsequent boot to succeed.

Fixes: 401d1c4f5d ("common: Drop asm/global_data.h from common header")
Signed-off-by: Matt Merhar <mattmerhar@protonmail.com>
2 years agopart: Add check for NULL dev_part_str
Sean Anderson [Sat, 15 May 2021 18:13:54 +0000 (14:13 -0400)]
part: Add check for NULL dev_part_str

Some callers (e.g. cmd/fs.c) of fs_set_blk_dev may use a NULL dev_part_str.
While blk_get_device_part_str handles this fine,
part_get_info_by_dev_and_name does not. This fixes commands crashing when
implicitly using bootdevice.

The unit test has also been updated to set bootdevice to a known value and
to restore it after we are done.

Fixes: 7194527b6a ("cmd: fs: Use part_get_info_by_dev_and_name_or_num to parse partitions")
Reported-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Sean Anderson <seanga2@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agopinctrl: single: Fix probe failure getting register area size
Vignesh Raghavendra [Fri, 7 May 2021 09:10:34 +0000 (14:40 +0530)]
pinctrl: single: Fix probe failure getting register area size

If reg property of pinctrl-single node requires address translation then
probe fails with following message:

single-pinctrl pinctrl@4301c000: failed to get base register size

This is because driver uses dev_read_addr_size() to get size which also
tries to fetch untranslated addr and fails.
Fix this by using dev_read_addr_size_index() which takes care of address
translation and also makes following dev_read_addr() call redundant.

This fixes Ethernet failures on TI's AM654 based EVMs due to lack of
pinmux configuration.

Fixes: 9fd8a430f3 ("pinctrl: single: get register area size by device API")
Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
2 years agoconfigs: Resync with savedefconfig
Tom Rini [Tue, 25 May 2021 15:58:50 +0000 (11:58 -0400)]
configs: Resync with savedefconfig

Rsync all defconfig files using moveconfig.py

Signed-off-by: Tom Rini <trini@konsulko.com>
2 years agonet: Remove ne2000 driver
Tom Rini [Tue, 25 May 2021 15:54:12 +0000 (11:54 -0400)]
net: Remove ne2000 driver

With the last user of this driver removed, remove the driver.

Reviewed-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Signed-off-by: Tom Rini <trini@konsulko.com>
2 years agoMerge tag 'efi-2021-07-rc4' of https://source.denx.de/u-boot/custodians/u-boot-efi
Tom Rini [Tue, 25 May 2021 15:48:55 +0000 (11:48 -0400)]
Merge tag 'efi-2021-07-rc4' of https://source.denx.de/u-boot/custodians/u-boot-efi

Pull request for efi-2021-07-rc4

Documentation:

* correct mmc man-page

Bug fixes:

* reduce code size of efidebug command
* remove 31 character limit for file paths in efidebug command
* fix build warning in the TCG2 protocol implementation

2 years agoMerge tag 'mips-pull-2021-05-25' of https://source.denx.de/u-boot/custodians/u-boot...
Tom Rini [Tue, 25 May 2021 15:48:37 +0000 (11:48 -0400)]
Merge tag 'mips-pull-2021-05-25' of https://source.denx.de/u-boot/custodians/u-boot-mips

- MIPS: octeon: fix CFI flash setup
- MIPS: remove qemu_mips boards

2 years agodoc: update and fix Qemu MIPS documentation
Daniel Schwierzeck [Wed, 21 Apr 2021 19:13:14 +0000 (21:13 +0200)]
doc: update and fix Qemu MIPS documentation

Update description to use the MIPS Malta board for Qemu.

Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
2 years agoMIPS: remove deprecated qemu_mips board
Daniel Schwierzeck [Wed, 21 Apr 2021 19:13:13 +0000 (21:13 +0200)]
MIPS: remove deprecated qemu_mips board

Remove qemu_mips boards because DM migration doesn't make sense.
The board support for qemu_mips is already marked as deprecated
in Qemu in favour of the Malta board. Also qemu_mips support
has been removed from Linux a long time ago.

The official replacement is the Malta board. The same Malta U-Boot
image can be used with Qemu and on physical hardware.
All combinations of Big Endian and Little Endian as well as 32 bit
and 64 bit are supported.

Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
2 years agomips: octeon: octeon_ebb7304_defconfig: Fix CFI flash setup
Stefan Roese [Mon, 26 Apr 2021 14:43:18 +0000 (16:43 +0200)]
mips: octeon: octeon_ebb7304_defconfig: Fix CFI flash setup

This patch makes the necessary adjustments in the defconfig to fully
support the CFI flash on the Octeon EBB7304.

Signed-off-by: Stefan Roese <sr@denx.de>
Cc: Aaron Williams <awilliams@marvell.com>
Cc: Chandrakala Chavva <cchavva@marvell.com>
Cc: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
2 years agoefi_loader: Fix -Wextra warning for EFI TCG2
Ilias Apalodimas [Tue, 25 May 2021 11:35:31 +0000 (14:35 +0300)]
efi_loader: Fix -Wextra warning for EFI TCG2

Compiling with -Wextra produces a number of (harmless) warnings for the EFI
TCG2 protocol.  Let's fix those in case we ever enable the flag on the EFI
subsystem

Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
2 years agoefi_loader: simplify accessing variables
Heinrich Schuchardt [Mon, 24 May 2021 09:10:59 +0000 (11:10 +0200)]
efi_loader: simplify accessing variables

Use efi_get_variable_int() instead of EFI_CALL(RT->get_variable()).
Use efi_set_variable_int() instead of EFI_CALL(efi_set_variable()).

Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Tested-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
2 years agoefi_loader: use efi_create_indexed_name()
Heinrich Schuchardt [Mon, 24 May 2021 08:57:03 +0000 (10:57 +0200)]
efi_loader: use efi_create_indexed_name()

Simplify the creation of indexed variable names like 'Boot0000' by using
function efi_create_indexed_name().

Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
2 years agoefi_loader: simplify show_efi_boot_order()
Heinrich Schuchardt [Mon, 24 May 2021 08:48:45 +0000 (10:48 +0200)]
efi_loader: simplify show_efi_boot_order()

To print a UTF-16 string use %ls instead of converting string to UTF-8.

Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
2 years agoefi_loader: simplify show_efi_boot_opt_data()
Heinrich Schuchardt [Mon, 24 May 2021 08:35:25 +0000 (10:35 +0200)]
efi_loader: simplify show_efi_boot_opt_data()

To print a UTF-16 string use %ls instead of converting string to UTF-8.

Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
2 years agoefi_loader: path length in efi_dp_from_name()
Heinrich Schuchardt [Tue, 25 May 2021 10:07:30 +0000 (12:07 +0200)]
efi_loader: path length in efi_dp_from_name()

Before this patch efi_dp_from_name() only accommodated a maximum file path
length of 31 characters. This leads to boot failures due to file name
truncation.

Allow arbitrary path lengths.

Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
2 years agoefi_loader: expose efi_image_parse() even if UEFI Secure Boot is disabled
Masahisa Kojima [Fri, 14 May 2021 00:53:36 +0000 (09:53 +0900)]
efi_loader: expose efi_image_parse() even if UEFI Secure Boot is disabled

This is preparation for PE/COFF measurement support.
PE/COFF image hash calculation is same in both
UEFI Secure Boot image verification and measurement in
measured boot. PE/COFF image parsing functions are
gathered into efi_image_loader.c, and exposed even if
UEFI Secure Boot is not enabled.

This commit also adds the EFI_SIGNATURE_SUPPORT option
to decide if efi_signature.c shall be compiled.

Signed-off-by: Masahisa Kojima <masahisa.kojima@linaro.org>
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
2 years agoefi_loader: remove weak efi_get_public_key_data()
Heinrich Schuchardt [Thu, 20 May 2021 02:08:11 +0000 (04:08 +0200)]
efi_loader: remove weak efi_get_public_key_data()

Configuring a system with CONFIG_EFI_CAPSULE_AUTHENTICATE=y but without a
method to retrieve the public key data is a misconfiguration. We should get
a build failure for it. Therefore remove the weak efi_get_public_key_data()
implementation.

Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
2 years agodoc: usage: modify mmc man-page
Jaehoon Chung [Mon, 10 May 2021 07:04:02 +0000 (16:04 +0900)]
doc: usage: modify mmc man-page

Modify mmc man-page.
- Change a description about mmc erase command.
- Add whitespace to distinguish.

Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
2 years agoPrepare v2021.07-rc3 v2021.07-rc3
Tom Rini [Tue, 25 May 2021 00:53:13 +0000 (20:53 -0400)]
Prepare v2021.07-rc3

Signed-off-by: Tom Rini <trini@konsulko.com>
2 years agoMerge branch '2021-05-24-add-lto-support'
Tom Rini [Mon, 24 May 2021 20:12:31 +0000 (16:12 -0400)]
Merge branch '2021-05-24-add-lto-support'

- Add LTO (link time optimization) support to the build system and
  enable it on a few boards.  This is an alternative to using
  -ffunction-sections/-fdata-sections and --gc-sections at link time to
  remove unused code.  This can result in notable savings, but needs
  testing on each platform before use as it can expose problems by
  optimizing away various functionally necessary calls.

2 years agostackprot: Make our test a bit more complex
Tom Rini [Mon, 24 May 2021 18:19:05 +0000 (14:19 -0400)]
stackprot: Make our test a bit more complex

With better compiler optimizations available, a compiler may see we do
nothing with our buffer after calling memset and omit the call, thus
causing us to not smash the stack.  Add a comment to explain why we now
also have a printf call, so that the test will pass as the memset will
not be omitted.

Reported-by: Marek Behún <marek.behun@nic.cz>
Signed-off-by: Tom Rini <trini@konsulko.com>
2 years agoARM: enable LTO for some boards
Marek Behún [Thu, 20 May 2021 11:24:25 +0000 (13:24 +0200)]
ARM: enable LTO for some boards

Enable LTO for some boards that were tested by people on U-Boot Mailing
List.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Tested-by: Adam Ford <aford173@gmail.com>
Tested-by: Pali Rohár <pali@kernel.org>
Tested-by: Tim Harvey <tharvey@gateworks.com>
2 years agoARM: don't use --gc-sections with LTO when using private libgcc
Marek Behún [Thu, 20 May 2021 11:24:24 +0000 (13:24 +0200)]
ARM: don't use --gc-sections with LTO when using private libgcc

When using LTO, we can throw away the --gc-sections flag, but only if
using private libgcc.

When using system's libgcc, --gc-sections is still needed, otherwise
linking will fail due to undefined references to libc's symbols.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agoARM: don't use -ffunction-sections/-fdata-sections with LTO build
Marek Behún [Thu, 20 May 2021 11:24:23 +0000 (13:24 +0200)]
ARM: don't use -ffunction-sections/-fdata-sections with LTO build

When building with LTO, using -ffunction-sections/-fdata-sections is not
useful anymore.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agoARM: make LTO available
Marek Behún [Thu, 20 May 2021 11:24:22 +0000 (13:24 +0200)]
ARM: make LTO available

Make LTO available for ARM architecture.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agoata: ahci: fix ahci_link_up() type mismatch for LTO
Marek Behún [Thu, 20 May 2021 11:24:21 +0000 (13:24 +0200)]
ata: ahci: fix ahci_link_up() type mismatch for LTO

When building highbank_defconfig with LTO, the compiler complains about
type mismatch of function ahci_link_up().

The third parameter of this function is of type u8 in
drivers/ata/ahci.c, but of type int in board/highbank/ahci.c.

There is no reason in using u8, and the code using this function
actually passes an int variable into the function (so it is implicitly
converted to u8).

Change the type of this parameter to int in drivers/ata/ahci.c.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
2 years agoarmv8: SPL: discard relocation information
Marek Behún [Thu, 20 May 2021 11:24:20 +0000 (13:24 +0200)]
armv8: SPL: discard relocation information

For some reason when building SPL for ARMv8 with LTO, the relocation
information is not discarded.

Discard it explicitly in the linker script.

This fixes LTO build for imx8mm_venice_defconfig.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agoARM: omap3: fix LTO for DM3730 (and possibly other omap3 boards)
Marek Behún [Thu, 20 May 2021 11:24:19 +0000 (13:24 +0200)]
ARM: omap3: fix LTO for DM3730 (and possibly other omap3 boards)

Adam Ford says that DM3730 needs board.c compiled without LTO flags.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Tested-by: Adam Ford <aford173@gmail.com>
2 years agoARM: fix LTO for rockchip and samsung
Marek Behún [Thu, 20 May 2021 11:24:18 +0000 (13:24 +0200)]
ARM: fix LTO for rockchip and samsung

When building with LTO, the compiler complains about type mismatch of
function usb_gadget_handle_interrupts(). This function is defined
without parameters in files
  arch/arm/mach-rockchip/board.c
  board/samsung/common/exynos5-dt.c
but it should have one parameter, int index.

Fix this.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
2 years agoARM: fix LTO for seaboard
Marek Behún [Thu, 20 May 2021 11:24:17 +0000 (13:24 +0200)]
ARM: fix LTO for seaboard

When seaboard_defconfig is compiled with LTO, the compiler complains
about some instructions not being supported in ARM mode.

This is caused by arch/arm/mach-tegra/tegra20/warmboot_avp.c having
different CFLAGS declared in Makefile. This file needs to be compiled
without LTO.

Fix this by removing -flto for this file.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agoARM: imx8m: fix imx_eqos_txclk_set_rate() type mismatch for LTO
Marek Behún [Thu, 20 May 2021 11:24:16 +0000 (13:24 +0200)]
ARM: imx8m: fix imx_eqos_txclk_set_rate() type mismatch for LTO

When building imx8mp_evk_defconfig with LTO, the compiler complains
about type mismatch of function imx_eqos_txclk_set_rate() in file
  drivers/net/dwc_eth_qos.c:845:12
which contains a weak definition of this function, vs file
  arch/arm/mach-imx/imx8m/clock_imx8mm.c
which contains an implementation.

Change the type of this function in the implementation to fix this.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
2 years agoARM: kona: fix clk_bsc_enable() type mismatch for LTO
Marek Behún [Thu, 20 May 2021 11:24:15 +0000 (13:24 +0200)]
ARM: kona: fix clk_bsc_enable() type mismatch for LTO

When building with LTO, the compiler complains about type mismatch of
function clk_bsc_enable() in file:
  arch/arm/cpu/armv7/kona-common/clk-stubs.c
vs other files that define or use this function:
  warning: type of ‘clk_bsc_enable’ does not match original declaration.

Change the type of this function to that of the other usages.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
2 years agoARM: fix LTO for keystone
Marek Behún [Thu, 20 May 2021 11:24:14 +0000 (13:24 +0200)]
ARM: fix LTO for keystone

When building keystone with LTO the compiler complains:
  Error: selected processor does not support `smc #0' in Thumb mode

Fix this by removing -flto for the file implementing these SMC calls.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
2 years agoARM: fix LTO for apf27
Marek Behún [Thu, 20 May 2021 11:24:13 +0000 (13:24 +0200)]
ARM: fix LTO for apf27

When apf27_defconfig is built with LTO, linking complains about
undefined reference to `nand_boot`. This is because it is referenced
from inline assembly. Make it visible.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
2 years agoARM: fix LTO for imx28_xea
Marek Behún [Thu, 20 May 2021 11:24:12 +0000 (13:24 +0200)]
ARM: fix LTO for imx28_xea

When imx28_xea_defconfig is built with LTO, the compiler complains about
the two different declarations of _start:
   include/asm-generic/sections.h    as  extern void _start(void);
   arch/arm/cpu/arm926ejs/mxs/mxs.c  as  extern uint32_t _start;

Fix this.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
2 years agoARM: fix LTO build for some thumb-interwork cases
Marek Behún [Thu, 20 May 2021 11:24:11 +0000 (13:24 +0200)]
ARM: fix LTO build for some thumb-interwork cases

Fix LTO build for some thumb-interwork usecases (such as for
da850evm_defconfig), where inline assmebly such as
  mrc p15,0,r2,c1,c0,0
causes the compiler to fail during LTO linking with
  Error: selected processor does not support `mrc p15,0,r2,c1,c0,0'
         in Thumb mode

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agoARM: make gd a function call for LTO and set via set_gd()
Marek Behún [Thu, 20 May 2021 11:24:10 +0000 (13:24 +0200)]
ARM: make gd a function call for LTO and set via set_gd()

On ARM, the gd pointer is stored in registers r9 / x18. For this the
-ffixed-r9 / -ffixed-x18 flag is used when compiling, but using global
register variables causes errors when building with LTO, and these
errors are very difficult to overcome.

Richard Biener says [1]:
  Note that global register vars shouldn't be used with LTO and if they
  are restricted to just a few compilation units the recommended fix is
  to build those CUs without -flto.

We cannot do this for U-Boot since all CUs use -ffixed-reg flag.

It seems that with LTO we could in fact store the gd pointer differently
and gain performance or size benefit by allowing the compiler to use
r9 / x18. But this would need more work.

So for now, when building with LTO, go the clang way, and instead of
declaring gd a global register variable, we make it a function call via
macro.

[1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68384

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agoARM: global_data: make set_gd() work for armv5 and armv6
Marek Behún [Thu, 20 May 2021 11:24:09 +0000 (13:24 +0200)]
ARM: global_data: make set_gd() work for armv5 and armv6

The Thumb instruction `ldr` is able to move high registers only from
armv7. For armv5 and armv6 we have to use `mov`.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agosandbox: enable LTO by default
Marek Behún [Thu, 20 May 2021 11:24:08 +0000 (13:24 +0200)]
sandbox: enable LTO by default

Build sandbox targets with LTO by default.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
2 years agosandbox: make LTO available
Marek Behún [Thu, 20 May 2021 11:24:07 +0000 (13:24 +0200)]
sandbox: make LTO available

Make LTO available for sandbox architecture.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
2 years agosandbox: use sections instead of symbols for getopt array boundaries
Marek Behún [Thu, 20 May 2021 11:24:06 +0000 (13:24 +0200)]
sandbox: use sections instead of symbols for getopt array boundaries

In style of linked lists, instead of declaring symbols for boundaries
of getopt options array in the linker script, declare corresponding
sections and retrieve the boundaries via static inline functions.

Without this clang's LTO produces binary without any getopt options,
because for some reason it thinks that array is empty (start and end
symbols are at the same address).

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agosandbox: errno: avoid conflict with libc's errno
Marek Behún [Thu, 20 May 2021 11:24:05 +0000 (13:24 +0200)]
sandbox: errno: avoid conflict with libc's errno

When building with LTO, the system libc's `errno` variable used in
arch/sandbox/cpu/os.c conflicts with U-Boot's `errno` (defined in
lib/errno.c) with the following error:
 .../ld: errno@@GLIBC_PRIVATE: TLS definition in /lib64/libc.so.6
         section .tbss mismatches non-TLS reference in
 /tmp/u-boot.EQlEXz.ltrans0.ltrans.o

To avoid this conflict use different asm label for this variable when
CONFIG_SANDBOX is enabled.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agobuild: link with --build-id=none
Marek Behún [Thu, 20 May 2021 11:24:04 +0000 (13:24 +0200)]
build: link with --build-id=none

Some toolchains are compiled so that they pass a --build-id=something
parameter to the linker implicitly.

This causes U-Boot LTO linking to fail with something like:
  ld: section .note.gnu.build-id LMA ... overlaps section .text LMA ...
because U-Boot's link scripts do not currently handle .note.gnu.build-id
section.

Fix this by explicitly disabling build-id.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agobuild: support building with Link Time Optimizations
Marek Behún [Thu, 20 May 2021 11:24:03 +0000 (13:24 +0200)]
build: support building with Link Time Optimizations

Add plumbing for building U-Boot with Link Time Optimizations.

When building with LTO, $(PLATFORM_LIBS) has to be in --whole-archive /
--no-whole-archive group, otherwise some functions declared in assembly
may not be resolved and linking may fail.

Note: clang may throw away linker list symbols it thinks are unused when
compiling with LTO. To force these symbols to be included, we refer to
them via the __ADDRESSABLE macro in a C file generated from compiled
built-in.o files before linking.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agobuild: use thin archives instead of incremental linking
Marek Behún [Thu, 20 May 2021 11:24:02 +0000 (13:24 +0200)]
build: use thin archives instead of incremental linking

Currently we use incremental linking (ld -r) to link several object
files from one directory into one built-in.o object file containing the
linked code from that directory (and its subdirectories).

Linux has, some time ago, moved to thin archives instead.

Thin archives are archives (.a) that do not really contain the object
files, only references to them.

Using thin archives instead of incremental linking
- saves disk space
- apparently works better with dead code elimination
- makes things easier for LTO

The third point is the important one for us. With incremental linking
there are several options how to do LTO, and that would unnecessarily
complicate things.

We have to use the --whole-archive/--no-whole-archive linking option
instead of --start-group/--end-group, otherwise linking may fail because
of unresolved symbols, or the resulting binary will be unusable.

We also need to use the P flag for ar, otherwise final linking may fail.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agoMakefile, Makefile.spl: cosmetic change
Marek Behún [Thu, 20 May 2021 11:24:01 +0000 (13:24 +0200)]
Makefile, Makefile.spl: cosmetic change

Indent the linking commands so that they look cosmetically better.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
2 years agolib: crc32: put the crc_table variable into efi_runtime_rodata section
Marek Behún [Thu, 20 May 2021 11:24:00 +0000 (13:24 +0200)]
lib: crc32: put the crc_table variable into efi_runtime_rodata section

When compiling with LTO, the compiler fails with an error saying that
`crc_table` causes a section type conflict with `efi_var_buf`.

This is because both are declared to be in the same section (via macro
`__efi_runtime_data`), but one is const while the other is not.

Put this variable into the section .rodata.efi_runtime, instead of
.data.efi_runtime, via macro __efi_runtime_rodata.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Marek Vasut <marex@denx.de>
Reviewed-by: Heinrich Schuchardt <xypron.gpk@gmx.de>
2 years agoefi_selftest: compiler flags for efi_selftest_miniapp_exception.o
Marek Behún [Thu, 20 May 2021 11:23:59 +0000 (13:23 +0200)]
efi_selftest: compiler flags for efi_selftest_miniapp_exception.o

Add $(CFLAGS_EFI) and remove $(CFLAGS_NON_EFI) for
efi_selftest_miniapp_exception.o.

The removal is needed when compiling with LTO - this object file needs
to be compiled without -flto.

The adding is for consistency with other miniapps.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
2 years agoefi_loader: add macro for const EFI runtime data
Marek Behún [Thu, 20 May 2021 11:23:58 +0000 (13:23 +0200)]
efi_loader: add macro for const EFI runtime data

Add macro __efi_runtime_rodata, for const variables with similar purpose
as those using __efi_runtime_data.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
2 years agoefi_loader: add Sphinx doc for __efi_runtime and __efi_runtime_data
Marek Behún [Thu, 20 May 2021 11:23:57 +0000 (13:23 +0200)]
efi_loader: add Sphinx doc for __efi_runtime and __efi_runtime_data

Document the macros __efi_runtime and __efi_runtime_data in Sphinx
style.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
2 years agoefi_loader: fix warning when linking with LTO
Marek Behún [Thu, 20 May 2021 11:23:56 +0000 (13:23 +0200)]
efi_loader: fix warning when linking with LTO

When linking with LTO, the compiler complains about type mismatch of
variables `__efi_runtime_start`, `__efi_runtime_stop`,
`__efi_runtime_rel_start` and `__efi_runtime_rel_stop`:

 include/efi_loader.h:218:21: warning: type of ‘__efi_runtime_start’
                                       does not match original
                                       declaration [-Wlto-type-mismatch]
    218 | extern unsigned int __efi_runtime_start, __efi_runtime_stop;
        |                     ^
  arch/sandbox/lib/sections.c:7:6: note: ‘__efi_runtime_start’ was
                                         previously declared here
      7 | char __efi_runtime_start[0] __attribute__((section(".__efi_run
        |      ^

Change the type to char[] in include/efi_loader.h.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>