Create the handle of loaded images and the EFI_LOADED_IMAGE_PROTOCOL
inside efi_setup_loaded_image(). Do not use local variables.
Currently we expect the loaded image handle to point to the loaded image
protocol. Additionally we have appended private fields to the protocol.
With the patch the handle points to a loaded image object and the private
fields are added here. This matches how we handle the net and the gop
object.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
When running in EL3 mode on AArch64, we have to first drop to EL2
to execute a UEFI payload. When dropping down, the arguments to
the entry point have to stay identical to the ones for normal entry
though.
In commit ea54ad5928 ("efi_loader: pass handle of loaded image")
we incorrectly changed that logic and had the el3 entry path diverge.
Fix it up by syncing it back to what it's supposed to be.
Fixes: ea54ad5928 ("efi_loader: pass handle of loaded image")
Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Mark Kettenis <kettenis@openbsd.org>
Currently we assign a lot of protocols to loaded images though
these protocols are not related to them. Instead they should be
installed on a separate handle. Via the device path it is the
parent to the devices like the network adapter.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
efi_set_bootdev() may be called repeatedly.
Free the memory allocated for device paths in previous calls.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
Do not use a local variable for the handle backing the memory device path.
Adjust relate comments.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
The number of bytes in an utf-8 string is an upper limit for the number of
words in the equivalent utf-16 string. In so far the inumbant coding works
correctly. For non-ASCII characters the utf-16 string is shorter. With the
patch only the necessary buffer size is allocated for the load options.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
At present this function takes a pointer as its argument, then passes this
to efi_allocate_pages(), which actually takes an address. It uses casts,
which are not supported on sandbox.
Also the function calculates the FDT size rounded up to the neared EFI
page size, then its caller recalculates the size and adds a bit more to
it.
This function is much better written as something that works with
addresses only, and returns both the address and the size of the relocated
FDT.
Also, copy_fdt() returns NULL on error, but really should propagate the
error from efi_allocate_pages(). To do this it needs to return an
efi_status_t, not a void *.
Update the code in this way, so that it is easier to follow, and also
supports sandbox.
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Sandbox only has 128MB of memory so we cannot relocate the device tree up
to start at 128MB. Use 127MB instead, which should be safe.
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Remove unused function efi_get_time_init().
Initialization of the RTC has to be done in board bring up not in the EFI
subsystem.
There is no RTC device in the UEFI spec. The RTC is only accessed through
the runtime services.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
For the boot and runtime services tables and for the system table the
crc32 has to be set in the header.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
This function currently returns an error code, but never uses it. There is
no function comment so it is not obvious why. Presuambly the error is not
important.
Update the function to explain its purpose and why it ignores the error.
Drop the useful error return value.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
This is a bit confusing at present since it adds 4KB to the pointer, then
rounds it up. It looks like a bug, but is not.
Move the 4KB addition into a separate statement and expand the comment.
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
The bootefi command gets a few addresses as values passed in. In sandbox,
these values are in U-Boot address space, so we need to make sure we
explicitly call map_sysmem() on them to be able to access them.
Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Multiple EFI binaries may be executed in sequence. So if we already
are in non-secure mode after running the first one we should skip
the switching code since it no longer works once we're non-secure.
Signed-off-by: Mark Kettenis <kettenis@openbsd.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
If desired (and possible) switch into HYP mode or non-secure SVC mode
before calling the entry point of an EFI application. This allows
U-Boot to provide a usable PSCI implementation and makes it possible
to boot kernels into hypervisor mode using an EFI bootloader.
Based on diffs from Heinrich Schuchardt and Alexander Graf.
Signed-off-by: Mark Kettenis <kettenis@openbsd.org>
[agraf: Fix indentation]
Signed-off-by: Alexander Graf <agraf@suse.de>
ACPI tables can be passed via EFI configuration table to an EFI
application. This is only supported on x86 so far.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
When we boot using memdp (bootefi on an address without previous
load that populates the device path) then the memory device path
we pass in is not backed by any handle.
That can result in weird effects. For example grub gets very grumpy
about this inside the efi_net module and just loops endlessly.
So let's expose a simple handle that the memory device path is backed
on. That way any code that looks for the device the dp is on, finds
one.
Signed-off-by: Alexander Graf <agraf@suse.de>
Do not use anonymous constants when calling efi_allocage_pages.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
The UEFI spec mandates that unaligned memory access should be enabled if
supported by the CPU architecture.
This patch adds an empty weak function unaligned_access() that can be
overridden by an architecture specific routine.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
Remove a superfluous call to efi_init_obj_list() invoked by
'bootefi selftest'.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
As part of the main conversion a few files were missed. These files had
additional whitespace after the '*' and before the SPDX tag and my
previous regex was too strict. This time I did a grep for all SPDX tags
and then filtered out anything that matched the correct styles.
Fixes: 83d290c56f ("SPDX: Convert all of our single license tags to Linux Kernel style")
Reported-by: Heinrich Schuchardt <xypron.debian@gmx.de>
Signed-off-by: Tom Rini <trini@konsulko.com>
This reverts the parts of commit 3b3ea2c56e
where it changed the EFI dependency on NET.
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com>
Reviewed-by: Duncan Hare <dh@synoia.com>
With legacy boot (booti, bootz), people can declare memory regions as
reserved using device tree memory reservations. This feature is some
times used to indicate memory regions that should not be touched.
Since in a UEFI world, the DT memory reservations do not get honored,
let's copy them into the UEFI memory map so everyone has a coherent
view of the world and we give people the chance to add reservations
on demand.
Signed-off-by: Alexander Graf <agraf@suse.de>
The device path of the loaded image should be set to the partition
from which the image was loaded. This requires using the same logic as
the load command.
Without the patch the device path pointed to the whole disk after executing
load mmc 0: 0x43000000 FILE
and not to the boot partition from which the file was actually loaded.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
The second argument of the bootefi command should always be usable to
specify a device tree. This was missing for bootefi selftest and
bootefi hello.
Proper error handling is added.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
Up to now errors in the initialization of the EFI subsystems was not
checked.
If any initialization fails, leave the bootefi command.
We do not retry initialization because this would require to undo all prior
initalization steps.
Suggested-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
The return type of do_bootefi_exec() is efi_status_t. So in case
of an error we should always return an EFI status code.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
efi_init_obj_list() should be executed only once.
Rather than having the caller check this variable and the callee set it,
move all access to the variable inside the function. This reduces the
logic needed to call efi_init_obj_list().
Suggested-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
Thomas reported U-Boot failed to build host tools if libfdt-devel
package is installed because tools include libfdt headers from
/usr/include/ instead of using internal ones.
This commit moves the header code:
include/libfdt.h -> include/linux/libfdt.h
include/libfdt_env.h -> include/linux/libfdt_env.h
and replaces include directives:
#include <libfdt.h> -> #include <linux/libfdt.h>
#include <libfdt_env.h> -> #include <linux/libfdt_env.h>
Reported-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Enable networking command only when NET is enabled.
And remove selecting NET for CMD_NET
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Signed-off-by: Tom Rini <trini@konsulko.com>
The bootefi command is missing in the online help for
bootefi bootmgr.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
If 'bootefi hello' or 'bootefi selftest' can be executed depends on the
configuration.
If an invalid non-numeric 1st argument is passed to bootefi, e.g.
'bootefi hola', this string is converted to 0 and U-Boot jumps to
this typically invalid address.
With the patch the online help is shown instead.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
EFI calls are usually defined as asmlinkage. That means we pass all parameters
to functions via the stack x86_32.
On x86_64 however, we need to also stick to the MS ABI calling conventions,
which the EFIAPI define conveniently handles for us. Most EFI functions were
also marked with EFIAPI, except for the entry call.
So this patch adjusts all entry calls to use EFIAPI instead of the manual
asmlinkage attribute.
While at it, we also change the prototype of the entry point to return
efi_status_t instead of ulong, as this is the correct prototype definition.
Signed-off-by: Alexander Graf <agraf@suse.de>
---
v1 -> v2:
- Use efi_status_t in all occurences
This patch provides
* a uclass for EFI drivers
* a EFI driver for block devices
For each EFI driver the uclass
* creates a handle
* adds the driver binding protocol
The uclass provides the bind, start, and stop entry points for the driver
binding protocol.
In bind() and stop() it checks if the controller implements the protocol
supported by the EFI driver. In the start() function it calls the bind()
function of the EFI driver. In the stop() function it destroys the child
controllers.
The EFI block driver binds to controllers implementing the block io
protocol.
When the bind function of the EFI block driver is called it creates a
new U-Boot block device. It installs child handles for all partitions and
installs the simple file protocol on these.
The read and write functions of the EFI block driver delegate calls to the
controller that it is bound to.
A usage example is as following:
U-Boot loads the iPXE snp.efi executable. iPXE connects an iSCSI drive and
exposes a handle with the block IO protocol. It calls ConnectController.
Now the EFI block driver installs the partitions with the simple file
protocol.
iPXE uses the simple file protocol to load Grub or the Linux Kernel.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
[agraf: add comment on calloc len]
Signed-off-by: Alexander Graf <agraf@suse.de>
The device tree is needed at runtime. So we have to store it in
EFI_RUNTIME_SERVICES_DATA memory.
The UEFI spec recommends to store all configuration tables in
EFI_RUNTIME_SERVICES_DATA memory.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
We should consistently use the efi_handle_t typedef when
referring to handles.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
The handle of a loaded image is the value of the handle
member of the loaded image info object and not the
address of the loaded image info.
Reviewed-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
When trying to load an image from a non-existent USB key, U-Boot v2017.11
crashes on my x86 platform:
=> load usb 0:1 03000000 abc
General Protection
EIP: 0010:[<7b59030d>] EFLAGS: 00010286
Original EIP :[<fff4330d>]
...
This used to work in v2017.09. Testing has shown, that this bug was
introduced with patch 95c5553e [efi_loader: refactor boot device and
loaded_image handling].
This patch now checks if a valid "desc" is returned from blk_get_dev()
and only continues when "desc" is available. Resulting in this cmd
output (again):
=> load usb 0:1 03000000 abc
** Bad device usb 0 **
Signed-off-by: Stefan Roese <sr@denx.de>
Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
Use environment variable bootargs used as load options
for bootefi payloads.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
After executing bootefi selftest
* restore GD
* unlink the load image handle
* return 0 or 1 and not a truncated efi_status_t.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
Environment variable efi_selftest is passed as load options
to the selftest application. It is used to select a single
test to be executed.
The load options are an UTF8 string. Yet I decided to keep
the name propertiy of the tests as char[] to reduce code
size.
Special value 'list' displays a list of all available tests.
Tests get an on_request property. If this property is set
the tests are only executed if explicitly requested.
The invocation of efi_selftest is changed to reflect that
bootefi selftest with efi_selftest = 'list' will call the
Exit bootservice.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
Currently we pass bootefi_device_path and bootefi_image_path as
device and image path without initializing them. They may carry
values from previous calls to bootefi.
With the patch the variables are initialized valid dummy values.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
Where ulong or unsigned long are used to hold an EFI status
code we should consistenly use efi_status_t.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
The watchdog is initialized with a 5 minute timeout period.
It can be reset by SetWatchdogTimer.
It is stopped by ExitBoottimeServices.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
When we don't have a real device/image path, such as 'bootefi hello',
construct a mem-mapped device-path.
This fixes 'bootefi hello' after devicepath refactoring.
Fixes: 95c5553ea2 ("efi_loader: refactor boot device and loaded_image handling")
Signed-off-by: Rob Clark <robdclark@gmail.com>
Acked-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
efi_exit() already restores gd, so we shouldn't EFI_EXIT() on the
otherside of the longjmp().
Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
ad503ffe9c6 efi_loader: refactor boot device and loaded_image handling
leads to an error when building with CONFIG_CMD_BOOTEFI_SELFTEST=y
This patch fixes the problem.
Fixes: ad503ffe9c6 efi_loader: refactor boot device and loaded_image handling
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
Similar to a "real" UEFI implementation, the bootmgr looks at the
BootOrder and BootXXXX variables to try to find an EFI payload to load
and boot. This is added as a sub-command of bootefi.
The idea is that the distro bootcmd would first try loading a payload
via the bootmgr, and then if that fails (ie. first boot or corrupted
EFI variables) it would fallback to loading bootaa64.efi. (Which
would then load fallback.efi which would look for \EFI\*\boot.csv and
populate BootOrder and BootXXXX based on what it found.)
Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
Add EFI variable support, mapping to u-boot environment variables.
Variables are pretty important for setting up boot order, among other
things. If the board supports saveenv, then it will be called in
ExitBootServices() to persist variables set by the efi payload. (For
example, fallback.efi configuring BootOrder and BootXXXX load-option
variables.)
Variables are *not* currently exposed at runtime, post ExitBootServices.
On boards without a dedicated device for storage, which the loaded OS
is not trying to also use, this is rather tricky. One idea, at least
for boards that can persist RAM across reboot, is to keep a "journal"
of modified variables in RAM, and then turn halt into a reboot into
u-boot, plus store variables, plus halt. Whatever the solution, it
likely involves some per-board support.
Mapping between EFI variables and u-boot variables:
efi_$guid_$varname = {attributes}(type)value
For example:
efi_8be4df61-93ca-11d2-aa0d-00e098032b8c_OsIndicationsSupported=
"{ro,boot,run}(blob)0000000000000000"
efi_8be4df61-93ca-11d2-aa0d-00e098032b8c_BootOrder=
"(blob)00010000"
The attributes are a comma separated list of these possible
attributes:
+ ro - read-only
+ boot - boot-services access
+ run - runtime access
NOTE: with current implementation, no variables are available after
ExitBootServices, and all are persisted (if possible).
If not specified, the attributes default to "{boot}".
The required type is one of:
+ utf8 - raw utf8 string
+ blob - arbitrary length hex string
Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
Get rid of the hacky fake boot-device and duplicate device-path
constructing (which needs to match what efi_disk and efi_net do).
Instead convert over to use efi_device_path helpers to construct
device-paths, and use that to look up the actual boot device.
Also, extract out a helper to plug things in properly to the
loaded_image. In a following patch we'll want to re-use this in
efi_load_image() to handle the case of loading an image from a
file_path.
Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
