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>
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>
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>
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.
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>
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>
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>
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>
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>
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>
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>
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.
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.
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>
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>
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>
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>
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>
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>
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>
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.
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>
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.
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>
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 [ced299d0, ced2a9d0]
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>
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>
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>
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>
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
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>
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
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>
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>
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>
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.
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>
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>
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>
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>
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>
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>
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.
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>
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>
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.
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>
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>
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>
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>
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>
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>
Marek Behún [Thu, 20 May 2021 11:23:55 +0000 (13:23 +0200)]
string: make memcpy(), memset(), memcmp() and memmove() visible for LTO
It seems that sometimes (happening on ARM64, for example with
turris_mox_defconfig) GCC, when linking with LTO, changes the symbol
names of some functions, for example lib/string.c's memcpy() function to
memcpy.isra.0.
This is a problem however when GCC for a code such as this:
struct some_struct *info = get_some_struct();
struct some struct tmpinfo;
tmpinfo = *info;
emits a call to memcpy() by builtin behaviour, to copy *info to tmpinfo.
This then results in the following linking error:
.../lz4.c:93: undefined reference to `memcpy'
.../uuid.c:206: more undefined references to `memcpy' follow
GCC's documentation says this about -nodefaultlibs option:
The compiler may generate calls to "memcmp", "memset", "memcpy" and
"memmove". These entries are usually resolved by entries in libc.
These entry points should be supplied through some other mechanism
when this option is specified.
Make these functions visible by using the __used macro to avoid this
error.
Signed-off-by: Marek Behún <marek.behun@nic.cz> Reviewed-by: Simon Glass <sjg@chromium.org>
Marek Behún [Thu, 20 May 2021 11:23:54 +0000 (13:23 +0200)]
test/py: improve regular expression for ut subtest symbol matcher
Improve the regular expression that matches unittest symbols in
u-boot.sym.
Currently we do not enforce no prefix in symbol string, but with the
soon to come change in linker lists declaring lists and entries with the
__ADDRESSABLE macro (because of LTO), the symbol file will contain for
every symbol of the form
_u_boot_list_2_ut_X_2_Y
also symbol
__UNIQUE_ID___addressable__u_boot_list_2_ut_X_2_YN,
(where N at the end is some number).
In order to avoid matching these additional symbols, ensure that the
character before "_u_boot_list_2_ut" is not a symbol name character.
Signed-off-by: Marek Behún <marek.behun@nic.cz> Reviewed-by: Simon Glass <sjg@chromium.org>
Marek Behún [Thu, 20 May 2021 11:23:52 +0000 (13:23 +0200)]
treewide: Convert macro and uses of __section(foo) to __section("foo")
This commit does the same thing as Linux commit 33def8498fdd.
Use a more generic form for __section that requires quotes to avoid
complications with clang and gcc differences.
Remove the quote operator # from compiler_attributes.h __section macro.
Convert all unquoted __section(foo) uses to quoted __section("foo").
Also convert __attribute__((section("foo"))) uses to __section("foo")
even if the __attribute__ has multiple list entry forms.
Signed-off-by: Marek Behún <marek.behun@nic.cz> Reviewed-by: Bin Meng <bmeng.cn@gmail.com> Reviewed-by: Simon Glass <sjg@chromium.org>
projects / u-boot.git / log