efi: Improvements to U-Boot running on top of UEFI
Subsystems (5)
v2 cover efi: Improvements to U-Boot running on top of UEFI
At present U-Boot can be built as an EFI app, but it is really just for testing, with very few features. Instead, the payload build is used for booting on top of UEFI, where U-Boot takes over the machine immediately and supplies its own drivers.
But the app could be made more useful.
This series provides access to EFI block devices and the video console within U-Boot. It also restructures the app to make it possible to boot a kernel, although further work is needed to implement this.
This can be thought of as making U-Boot perform a little like 'grub', in that it runs purely based on UEFI-provided services.
Of course a lot more work is required, but this is a start.
Note: It would be very useful to include qemu tests of the app and stub in CI.
This is available at u-boot-dm/efi-working
Changes in v2:
- Add MAINTAINERS entry
- Add a better boot command too
- Add a note that EFI_GRAPHICS_OUTPUT_PROTOCOL is used
- Add a sentence about what the patch does
- Add a work-around to avoid a toolchain crash
- Add new patch to drop the OF_EMBED warning for EFI
- Add new patch to enable DM_ETH for the app
- Add new patch to support the efi command in the app
- Add new patch with maintainer entry
- Add support for creating a partition table with a filesystem inside
- Add support for running qemu with just a serial console (no display)
- Drop mention of partitions from the commit message
- Expand the commit message to make things clearer
- Fix 'as' typo
- Mention why this patch is included in this series
- Put this patch first so that it is separated from the rest of the series
- Show the correct interface type with 'part list'
- Update commit message to mention that these function are x86-specific
- Update commit message to mention why the data structure is like it is
- Update commit message to mention why ulong is used
- Update documentation
- Update the commit message to explain things better
- Use log_info() instead of printf()
Simon Glass (39):
RFC: efi: Drop code that doesn't work with driver model
efi: Add a separate maintainer entry for the app
x86: Keep symbol information in u-boot ELF file
x86: Create a new header for EFI
x86: Show some EFI info with the bdinfo command
x86: Tidy up global_data pointer for 64-bit
efi: Add a script for building and testing U-Boot on UEFI
efi: Enable DM_ETH for the app
efi: Drop the OF_EMBED warning for EFI
x86: Create a 32/64-bit selection for the app
efi: Create a 64-bit app
x86: Don't duplicate global_ptr in 64-bit EFI app
efi: Add a way to obtain boot services in the app
efi: Add video support to the app
efi: Add EFI uclass for media
efi: Add a media/block driver for EFI block devices
efi: Locate all block devices in the app
patman: Use a ValueError exception if tools.Run() fails
binman: Report an error if test files fail to compile
binman: Support reading the offset of an ELF-file symbol
binman: Allow timeout to occur in the image or its section
binman: Tidy up comments on _DoTestFile()
binman: Support updating the dtb in an ELF file
efi: serial: Support arrow keys
bloblist: Support allocating the bloblist
x86: Allow booting a kernel from the EFI app
x86: Don't process the kernel command line unless enabled
x86: efi: Add room for the binman definition in the dtb
efi: Add comments to struct efi_priv
efi: Fix ll_boot_init() operation with the app
efi: Add a few comments to the stub
efi: Share struct efi_priv between the app and stub code
efi: Move exit_boot_services into a function
efi: Check for failure when initing the app
efi: Mention that efi_info_get() is only used in the stub
efi: Show when allocated pages are used
efi: Allow easy selection of serial-only operation
efi: Update efi_get_next_mem_desc() to avoid needing a map
efi: Support the efi command in the app
MAINTAINERS | 14 +
Makefile | 10 +-
arch/sandbox/dts/test.dts | 4 +
arch/x86/config.mk | 4 +-
arch/x86/cpu/efi/payload.c | 21 +-
arch/x86/cpu/intel_common/Makefile | 2 +-
arch/x86/cpu/u-boot-64.lds | 2 +
arch/x86/cpu/x86_64/Makefile | 4 +
arch/x86/cpu/x86_64/cpu.c | 32 +-
arch/x86/cpu/x86_64/misc.c | 41 +++
arch/x86/dts/Makefile | 2 +-
arch/x86/dts/efi-x86_app.dts | 4 +
arch/x86/include/asm/efi.h | 39 +++
arch/x86/include/asm/global_data.h | 2 +
arch/x86/include/asm/zimage.h | 3 -
arch/x86/lib/Makefile | 1 +
arch/x86/lib/bdinfo.c | 22 ++
arch/x86/lib/bootm.c | 11 +-
arch/x86/lib/zimage.c | 14 +-
board/efi/Kconfig | 15 +-
board/efi/efi-x86_app/Kconfig | 6 +-
board/efi/efi-x86_app/MAINTAINERS | 11 +-
cmd/Makefile | 2 +-
cmd/efi.c | 76 +++--
common/Kconfig | 15 +-
common/bloblist.c | 16 +-
common/board_f.c | 8 +-
..._app_defconfig => efi-x86_app32_defconfig} | 3 +-
configs/efi-x86_app64_defconfig | 39 +++
disk/part.c | 5 +-
doc/develop/bloblist.rst | 16 +
doc/develop/uefi/u-boot_on_efi.rst | 74 ++++-
drivers/block/Kconfig | 33 +++
drivers/block/Makefile | 4 +
drivers/block/blk-uclass.c | 2 +-
drivers/block/efi-media-uclass.c | 15 +
drivers/block/efi_blk.c | 115 ++++++++
drivers/block/sb_efi_media.c | 20 ++
drivers/serial/serial_efi.c | 11 +-
drivers/video/Kconfig | 2 +-
drivers/video/efi.c | 45 ++-
include/configs/efi-x86_app.h | 31 +-
include/dm/uclass-id.h | 1 +
include/efi.h | 123 +++++++-
include/efi_api.h | 15 +
include/init.h | 2 +-
lib/efi/Kconfig | 34 ++-
lib/efi/efi.c | 106 +++++++
lib/efi/efi_app.c | 276 +++++++++++++++++-
lib/efi/efi_stub.c | 95 +++---
lib/efi_driver/Makefile | 2 +-
lib/efi_loader/Kconfig | 1 +
lib/efi_loader/efi_device_path.c | 96 ++----
lib/efi_loader/efi_disk.c | 48 ---
scripts/build-efi.sh | 188 ++++++++++++
test/dm/Makefile | 1 +
test/dm/efi_media.c | 24 ++
tools/binman/binman.rst | 36 +++
tools/binman/cmdline.py | 2 +
tools/binman/control.py | 11 +
tools/binman/elf.py | 74 ++++-
tools/binman/elf_test.py | 45 ++-
tools/binman/ftest.py | 91 +++++-
tools/binman/test/Makefile | 13 +-
tools/binman/test/bss_data.c | 2 +-
tools/binman/test/embed_data.c | 16 +
tools/binman/test/embed_data.lds | 23 ++
tools/binman/test/u_boot_binman_embed.c | 13 +
tools/binman/test/u_boot_binman_embed.lds | 29 ++
tools/binman/test/u_boot_binman_embed_sm.c | 13 +
tools/patman/tools.py | 2 +-
71 files changed, 1841 insertions(+), 337 deletions(-)
create mode 100644 arch/x86/cpu/x86_64/misc.c
create mode 100644 arch/x86/include/asm/efi.h
create mode 100644 arch/x86/lib/bdinfo.c
rename configs/{efi-x86_app_defconfig => efi-x86_app32_defconfig} (94%)
create mode 100644 configs/efi-x86_app64_defconfig
create mode 100644 drivers/block/efi-media-uclass.c
create mode 100644 drivers/block/efi_blk.c
create mode 100644 drivers/block/sb_efi_media.c
create mode 100755 scripts/build-efi.sh
create mode 100644 test/dm/efi_media.c
create mode 100644 tools/binman/test/embed_data.c
create mode 100644 tools/binman/test/embed_data.lds
create mode 100644 tools/binman/test/u_boot_binman_embed.c
create mode 100644 tools/binman/test/u_boot_binman_embed.lds
create mode 100644 tools/binman/test/u_boot_binman_embed_sm.cSignature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 1/39 RFC: efi: Drop code that doesn't work with driver model
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on DM_ETH also, to avoid needing to deal with this old code.
Boards which want EFI_LOADER should migrate to driver model first.
Note this patch is included to resolve the following build error:
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’
undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?
680 | ulong base = CONFIG_SYS_TEXT_BASE;
| ^~~~~~~~~~~~~~~~~~~~
| CONFIG_SYS_SRAM_BASE---
Changes in v2:
- Mention why this patch is included in this series
- Put this patch first so that it is separated from the rest of the series
lib/efi_driver/Makefile | 2 +- lib/efi_loader/Kconfig | 1 + lib/efi_loader/efi_device_path.c | 96 +++++++------------------------- lib/efi_loader/efi_disk.c | 48 ---------------- 4 files changed, 23 insertions(+), 124 deletions(-)
lib/efi_driver/Makefile +1 -1
@@ -6,6 +6,6 @@ # object inclusion implicitly depends on it obj-y += efi_uclass.o -ifeq ($(CONFIG_BLK)$(CONFIG_PARTITIONS),yy) +ifeq ($(CONFIG_PARTITIONS),y) obj-y += efi_block_device.o endif
lib/efi_loader/Kconfig +1 -0
@@ -11,6 +11,7 @@ config EFI_LOADER # We need EFI_STUB_32BIT to be set on x86_32 with EFI_STUB depends on !EFI_STUB || !X86 || X86_64 || EFI_STUB_32BIT depends on BLK + depends on DM_ETH || !NET default y if !ARM || SYS_CPU = armv7 || SYS_CPU = armv8 select LIB_UUID select PARTITION_UUIDS
lib/efi_loader/efi_device_path.c +21 -75
197 lines not on this page · raw message
lib/efi_loader/efi_disk.c +0 -48
61 lines not on this page · raw message
Signature · 1 line
2.33.0.685.g46640cef36-goog
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
Quoted 10 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
Quoted 8 lines
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’
undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?680 | ulong base = CONFIG_SYS_TEXT_BASE; | ^~~~~~~~~~~~~~~~~~~~ | CONFIG_SYS_SRAM_BASESigned-off-by Simon Glass <[email protected]>
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
Quoted 19 lines · Heinrich Schuchardt
Quoted 18 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’ undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?
680 | ulong base =CONFIG_SYS_TEXT_BASE;
| ^~~~~~~~~~~~~~~~~~~~
|CONFIG_SYS_SRAM_BASE
Quoted 4 lines
Quoted 1 line
Signed-off-by Simon Glass <[email protected]>Reviewed-by Heinrich Schuchardt <[email protected]>
How to deal with boards that need CONFIG_NET but do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.
Jan
Signature · 2 lines
Siemens AG, Technology Competence Center Embedded Linux
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
Quoted 22 lines · Jan Kiszka
Quoted 21 lines · Heinrich Schuchardt
Quoted 19 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’ undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?
680 | ulong base =CONFIG_SYS_TEXT_BASE;
| ^~~~~~~~~~~~~~~~~~~~
|CONFIG_SYS_SRAM_BASESigned-off-by Simon Glass <[email protected]>
Quoted 7 lines
Quoted 2 lines
Reviewed-by Heinrich Schuchardt <[email protected]>How to deal with boards that need
CONFIG_NETbut do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.
Yes, looking at the patch, this should probably use NETDEVICES and not NET as the test? Would that fix your problem, Jan?
Signature · 1 line
Tom
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
Hi Tom,
Quoted 27 lines · Tom Rini
Quoted 25 lines · Jan Kiszka
Quoted 23 lines · Heinrich Schuchardt
Quoted 11 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
Quoted 8 lines
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’
undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?680 | ulong base = CONFIG_SYS_TEXT_BASE; | ^~~~~~~~~~~~~~~~~~~~ | CONFIG_SYS_SRAM_BASESigned-off-by Simon Glass <[email protected]>Reviewed-by Heinrich Schuchardt <[email protected]>
Quoted 7 lines
Quoted 4 lines
How to deal with boards that need
CONFIG_NETbut do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.Yes, looking at the patch, this should probably use NETDEVICES and not NET as the test? Would that fix your problem, Jan?
But that selects DM_ETH. You can still enable DM_ETH when there are no devices, so I'm not sure what the problem is. Can you post an error?
Things should be migrated to DM_ETH by now (end of 2020).
Regards,
Simon
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
Quoted 35 lines · Simon Glass
Hi Tom,
Quoted 32 lines · Tom Rini
Quoted 29 lines · Jan Kiszka
Quoted 19 lines · Heinrich Schuchardt
Quoted 18 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’
undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?680 | ulong base = CONFIG_SYS_TEXT_BASE; | ^~~~~~~~~~~~~~~~~~~~ | CONFIG_SYS_SRAM_BASEQuoted 4 lines
Quoted 1 line
Signed-off-by Simon Glass <[email protected]>Reviewed-by Heinrich Schuchardt <[email protected]>How to deal with boards that need
CONFIG_NETbut do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.
Quoted 8 lines
Quoted 2 lines
Yes, looking at the patch, this should probably use NETDEVICES and not NET as the test? Would that fix your problem, Jan?
But that selects
DM_ETH. You can still enableDM_ETHwhen there are no devices, so I'm not sure what the problem is. Can you post an error?Things should be migrated to
DM_ETHby now (end of 2020).
I am enable DM_ETH now at board-level, just to make EFI_LOADER happy, but that feels wrong.
Jan
Signature · 2 lines
Siemens AG, Technology Competence Center Embedded Linux
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
Hi Jan,
Quoted 40 lines · Jan Kiszka
Quoted 38 lines · Simon Glass
Hi Tom,
Quoted 34 lines · Tom Rini
Quoted 22 lines · Jan Kiszka
Quoted 21 lines · Heinrich Schuchardt
Quoted 19 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’
undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?680 | ulong base = CONFIG_SYS_TEXT_BASE; | ^~~~~~~~~~~~~~~~~~~~ | CONFIG_SYS_SRAM_BASESigned-off-by Simon Glass <[email protected]>Quoted 7 lines
Quoted 2 lines
Reviewed-by Heinrich Schuchardt <[email protected]>How to deal with boards that need
CONFIG_NETbut do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.Yes, looking at the patch, this should probably use NETDEVICES and not NET as the test? Would that fix your problem, Jan?
Quoted 8 lines
Quoted 5 lines
But that selects
DM_ETH. You can still enableDM_ETHwhen there are no devices, so I'm not sure what the problem is. Can you post an error?Things should be migrated to
DM_ETHby now (end of 2020).I am enable
DM_ETHnow at board-level, just to makeEFI_LOADERhappy, but that feels wrong.
I agree, but I wonder if this might need some code refactoring. Perhaps EFI_LOADER_NET needs to be created, net functionality can be disabled?
Regards,
Simon
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
Quoted 49 lines · Simon Glass
Hi Jan,
Quoted 46 lines · Jan Kiszka
Quoted 43 lines · Simon Glass
Hi Tom,
Quoted 27 lines · Tom Rini
Quoted 25 lines · Jan Kiszka
Quoted 23 lines · Heinrich Schuchardt
Quoted 11 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
Quoted 8 lines
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’
undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?680 | ulong base = CONFIG_SYS_TEXT_BASE; | ^~~~~~~~~~~~~~~~~~~~ | CONFIG_SYS_SRAM_BASESigned-off-by Simon Glass <[email protected]>Reviewed-by Heinrich Schuchardt <[email protected]>Quoted 7 lines
Quoted 4 lines
How to deal with boards that need
CONFIG_NETbut do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.Yes, looking at the patch, this should probably use NETDEVICES and not NET as the test? Would that fix your problem, Jan?
But that selects
DM_ETH. You can still enableDM_ETHwhen there are no devices, so I'm not sure what the problem is. Can you post an error?Things should be migrated to
DM_ETHby now (end of 2020).
Quoted 6 lines
Quoted 2 lines
I am enable
DM_ETHnow at board-level, just to makeEFI_LOADERhappy, but that feels wrong.I agree, but I wonder if this might need some code refactoring. Perhaps
EFI_LOADER_NETneeds to be created, net functionality can be disabled?
I think we should be able to leverage existing flags already for this and not need a new one. The issue is that CONFIG_NET does not mean "we have network devices" but CONFIG_NETDEVICES does.
Signature · 1 line
Tom
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
Quoted 53 lines · Tom Rini
Quoted 52 lines · Simon Glass
Hi Jan,
Quoted 48 lines · Jan Kiszka
Quoted 35 lines · Simon Glass
Hi Tom,
Quoted 32 lines · Tom Rini
Quoted 29 lines · Jan Kiszka
Quoted 19 lines · Heinrich Schuchardt
Quoted 18 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’
undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?680 | ulong base = CONFIG_SYS_TEXT_BASE; | ^~~~~~~~~~~~~~~~~~~~ | CONFIG_SYS_SRAM_BASEQuoted 4 lines
Quoted 1 line
Signed-off-by Simon Glass <[email protected]>Reviewed-by Heinrich Schuchardt <[email protected]>How to deal with boards that need
CONFIG_NETbut do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.Quoted 8 lines
Quoted 2 lines
Yes, looking at the patch, this should probably use NETDEVICES and not NET as the test? Would that fix your problem, Jan?
But that selects
DM_ETH. You can still enableDM_ETHwhen there are no devices, so I'm not sure what the problem is. Can you post an error?Things should be migrated to
DM_ETHby now (end of 2020).I am enable
DM_ETHnow at board-level, just to makeEFI_LOADERhappy, but that feels wrong.
Quoted 8 lines
Quoted 3 lines
I agree, but I wonder if this might need some code refactoring. Perhaps
EFI_LOADER_NETneeds to be created, net functionality can be disabled?I think we should be able to leverage existing flags already for this and not need a new one. The issue is that
CONFIG_NETdoes not mean "we have network devices" butCONFIG_NETDEVICESdoes.
So, you mean fixing up the efi-loader code to use CONFIG_NETDEVICES, rather than CONFIG_NET? Then we could drop that Kconfig check.
Jan
Signature · 2 lines
Siemens AG, Technology Competence Center Embedded Linux
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
Quoted 58 lines · Jan Kiszka
Quoted 57 lines · Tom Rini
Quoted 55 lines · Simon Glass
Hi Jan,
Quoted 40 lines · Jan Kiszka
Quoted 38 lines · Simon Glass
Hi Tom,
Quoted 34 lines · Tom Rini
Quoted 22 lines · Jan Kiszka
Quoted 21 lines · Heinrich Schuchardt
Quoted 19 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’
undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?680 | ulong base = CONFIG_SYS_TEXT_BASE; | ^~~~~~~~~~~~~~~~~~~~ | CONFIG_SYS_SRAM_BASESigned-off-by Simon Glass <[email protected]>Quoted 7 lines
Quoted 2 lines
Reviewed-by Heinrich Schuchardt <[email protected]>How to deal with boards that need
CONFIG_NETbut do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.Yes, looking at the patch, this should probably use NETDEVICES and not NET as the test? Would that fix your problem, Jan?
Quoted 8 lines
Quoted 5 lines
But that selects
DM_ETH. You can still enableDM_ETHwhen there are no devices, so I'm not sure what the problem is. Can you post an error?Things should be migrated to
DM_ETHby now (end of 2020).I am enable
DM_ETHnow at board-level, just to makeEFI_LOADERhappy, but that feels wrong.I agree, but I wonder if this might need some code refactoring. Perhaps
EFI_LOADER_NETneeds to be created, net functionality can be disabled?
Quoted 6 lines
Quoted 3 lines
I think we should be able to leverage existing flags already for this and not need a new one. The issue is that
CONFIG_NETdoes not mean "we have network devices" butCONFIG_NETDEVICESdoes.So, you mean fixing up the efi-loader code to use
CONFIG_NETDEVICES, rather thanCONFIG_NET? Then we could drop thatKconfigcheck.
Yes.
Signature · 1 line
Tom
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
In reply to Jan KiszkaQuoted 22 lines · Jan Kiszka
Quoted 21 lines · Heinrich Schuchardt
Quoted 19 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’
undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?
680 | ulong base =CONFIG_SYS_TEXT_BASE;
| ^~~~~~~~~~~~~~~~~~~~
|CONFIG_SYS_SRAM_BASESigned-off-by Simon Glass <[email protected]>
Quoted 7 lines
Quoted 2 lines
Reviewed-by Heinrich Schuchardt <[email protected]>How to deal with boards that need
CONFIG_NETbut do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.
UEFI will not provide network without a DM driver. The migration period ended 2020.07.
make iot2050_defconfig set EFI_LOADER=y.
What does not work in UEFI?
Are you missing a way to call eth_get_dev() to trigger aforementioned devicetree setup?
Adding 'net list' to CONFIG_PREBOOT might do the trick.
Best regards
Heinrich
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
Quoted 34 lines · Heinrich Schuchardt
Quoted 22 lines · Jan Kiszka
Quoted 21 lines · Heinrich Schuchardt
Quoted 19 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’
undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?
680 | ulong base =CONFIG_SYS_TEXT_BASE;
| ^~~~~~~~~~~~~~~~~~~~
|CONFIG_SYS_SRAM_BASESigned-off-by Simon Glass <[email protected]>Quoted 7 lines
Quoted 2 lines
Reviewed-by Heinrich Schuchardt <[email protected]>How to deal with boards that need
CONFIG_NETbut do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.UEFI will not provide network without a DM driver. The migration period ended 2020.07.
Quoted 7 lines
make
iot2050_defconfigset EFI_LOADER=y.
What does not work in UEFI?Are you missing a way to call
eth_get_dev()to trigger aforementioned devicetree setup?Adding 'net list' to
CONFIG_PREBOOTmight do the trick.
I think, and this is why I suggested NETDEVICES, the issue is that there is no network driver in U-Boot, but we need to have networking support enabled in general, in order to perform device tree fixups for Linux where there is the network driver.
Signature · 1 line
Tom
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
Quoted 41 lines · Tom Rini
Quoted 40 lines · Heinrich Schuchardt
Quoted 22 lines · Jan Kiszka
Quoted 21 lines · Heinrich Schuchardt
Quoted 19 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’
undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?
680 | ulong base =CONFIG_SYS_TEXT_BASE;
| ^~~~~~~~~~~~~~~~~~~~
|CONFIG_SYS_SRAM_BASESigned-off-by Simon Glass <[email protected]>Quoted 7 lines
Quoted 2 lines
Reviewed-by Heinrich Schuchardt <[email protected]>How to deal with boards that need
CONFIG_NETbut do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.UEFI will not provide network without a DM driver. The migration period ended 2020.07.
make
iot2050_defconfigset EFI_LOADER=y.
What does not work in UEFI?Are you missing a way to call
eth_get_dev()to trigger aforementioned devicetree setup?
Quoted 7 lines
Quoted 1 line
Adding 'net list' to
CONFIG_PREBOOTmight do the trick.I think, and this is why I suggested NETDEVICES, the issue is that there is no network driver in U-Boot, but we need to have networking support enabled in general, in order to perform device tree fixups for Linux where there is the network driver.
Exactly.
Unfortunately, NETDEVICES does not do the trick because it selects DM_ETH anyway, thus you don't detect legacy network drivers this way.
Jan
Signature · 2 lines
Siemens AG, Technology Competence Center Embedded Linux
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
Quoted 50 lines · Jan Kiszka
Quoted 41 lines · Tom Rini
Quoted 40 lines · Heinrich Schuchardt
Quoted 22 lines · Jan Kiszka
Quoted 21 lines · Heinrich Schuchardt
Quoted 19 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’
undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?
680 | ulong base =CONFIG_SYS_TEXT_BASE;
| ^~~~~~~~~~~~~~~~~~~~
|CONFIG_SYS_SRAM_BASESigned-off-by Simon Glass <[email protected]>Quoted 7 lines
Quoted 2 lines
Reviewed-by Heinrich Schuchardt <[email protected]>How to deal with boards that need
CONFIG_NETbut do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.UEFI will not provide network without a DM driver. The migration period ended 2020.07.
make
iot2050_defconfigset EFI_LOADER=y.
What does not work in UEFI?Are you missing a way to call
eth_get_dev()to trigger aforementioned devicetree setup?Quoted 7 lines
Quoted 1 line
Adding 'net list' to
CONFIG_PREBOOTmight do the trick.I think, and this is why I suggested NETDEVICES, the issue is that there is no network driver in U-Boot, but we need to have networking support enabled in general, in order to perform device tree fixups for Linux where there is the network driver.
Quoted 4 lines
Exactly.
Unfortunately, NETDEVICES does not do the trick because it selects
DM_ETHanyway, thus you don't detect legacy network drivers this way.
Right, but does your platform enable NETDEVICES? It sounds like it shouldn't be since you don't have devices in U-Boot? Or something else?
Signature · 1 line
Tom
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
Quoted 51 lines · Tom Rini
Quoted 50 lines · Jan Kiszka
Quoted 48 lines · Tom Rini
Quoted 34 lines · Heinrich Schuchardt
Quoted 22 lines · Jan Kiszka
Quoted 21 lines · Heinrich Schuchardt
Quoted 19 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’
undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?
680 | ulong base =CONFIG_SYS_TEXT_BASE;
| ^~~~~~~~~~~~~~~~~~~~
|CONFIG_SYS_SRAM_BASESigned-off-by Simon Glass <[email protected]>Quoted 7 lines
Quoted 2 lines
Reviewed-by Heinrich Schuchardt <[email protected]>How to deal with boards that need
CONFIG_NETbut do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.UEFI will not provide network without a DM driver. The migration period ended 2020.07.
Quoted 7 lines
make
iot2050_defconfigset EFI_LOADER=y.
What does not work in UEFI?Are you missing a way to call
eth_get_dev()to trigger aforementioned devicetree setup?Adding 'net list' to
CONFIG_PREBOOTmight do the trick.I think, and this is why I suggested NETDEVICES, the issue is that there is no network driver in U-Boot, but we need to have networking support enabled in general, in order to perform device tree fixups for Linux where there is the network driver.
Quoted 8 lines
Quoted 4 lines
Exactly.
Unfortunately, NETDEVICES does not do the trick because it selects
DM_ETHanyway, thus you don't detect legacy network drivers this way.Right, but does your platform enable NETDEVICES? It sounds like it shouldn't be since you don't have devices in U-Boot? Or something else?
NETDEVICES is off, obviously.
Jan
Signature · 2 lines
Siemens AG, Technology Competence Center Embedded Linux
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
Quoted 54 lines · Jan Kiszka
Quoted 53 lines · Tom Rini
Quoted 52 lines · Jan Kiszka
Quoted 41 lines · Tom Rini
Quoted 40 lines · Heinrich Schuchardt
Quoted 22 lines · Jan Kiszka
Quoted 21 lines · Heinrich Schuchardt
Quoted 19 lines · Simon Glass
This code should never have been added as it builds a new feature on top of legacy code. This has already been improved with the dependency on BLK.
Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’
undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?
680 | ulong base =CONFIG_SYS_TEXT_BASE;
| ^~~~~~~~~~~~~~~~~~~~
|CONFIG_SYS_SRAM_BASESigned-off-by Simon Glass <[email protected]>Quoted 7 lines
Quoted 2 lines
Reviewed-by Heinrich Schuchardt <[email protected]>How to deal with boards that need
CONFIG_NETbut do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.UEFI will not provide network without a DM driver. The migration period ended 2020.07.
make
iot2050_defconfigset EFI_LOADER=y.
What does not work in UEFI?Are you missing a way to call
eth_get_dev()to trigger aforementioned devicetree setup?Quoted 7 lines
Quoted 1 line
Adding 'net list' to
CONFIG_PREBOOTmight do the trick.I think, and this is why I suggested NETDEVICES, the issue is that there is no network driver in U-Boot, but we need to have networking support enabled in general, in order to perform device tree fixups for Linux where there is the network driver.
Exactly.
Quoted 8 lines
Quoted 6 lines
Quoted 2 lines
Unfortunately, NETDEVICES does not do the trick because it selects
DM_ETHanyway, thus you don't detect legacy network drivers this way.Right, but does your platform enable NETDEVICES? It sounds like it shouldn't be since you don't have devices in U-Boot? Or something else?
NETDEVICES is off, obviously.
Right, so, this patch should make the EFI code test for NETDEVICES and not NET, and then you wouldn't need to hack in something on your platform I think.
Signature · 1 line
Tom
Re: [PATCH v2 01/39] RFC: efi: Drop code that doesn't work with driver model
In reply to Heinrich SchuchardtQuoted 27 lines · Heinrich Schuchardt
Quoted 26 lines · Jan Kiszka
Quoted 24 lines · Heinrich Schuchardt
Quoted 12 lines · Simon Glass
This code should never have been added as it builds a new feature on top
of legacy code. This has already been improved with the dependency on BLK.Add a dependency on
DM_ETHalso, to avoid needing to deal with this old code.Boards which want
EFI_LOADERshould migrate to driver model first.Note this patch is included to resolve the following build error:
Quoted 8 lines
lib/efi_loader/efi_runtime.c:680:16: error: ‘CONFIG_SYS_TEXT_BASE’ undeclared (first use in this function); did you mean ‘CONFIG_SYS_SRAM_BASE’?
680 | ulong base =CONFIG_SYS_TEXT_BASE;
| ^~~~~~~~~~~~~~~~~~~~
|CONFIG_SYS_SRAM_BASESigned-off-by Simon Glass <[email protected]>Reviewed-by Heinrich Schuchardt <[email protected]>
Quoted 7 lines
Quoted 4 lines
How to deal with boards that need
CONFIG_NETbut do not actually implement any driver (yet)? This now broke UEFI for the IOT2050 which needs NET for network-related device tree setup (see also [1]) and enforces a local hack for us.UEFI will not provide network without a DM driver. The migration period ended 2020.07.
We are not interested in networking in UEFI so far. We need other features from CONFIG_NET, namely mac address transfer into the DT.
Quoted 3 lines
make
iot2050_defconfigset EFI_LOADER=y.
What does not work in UEFI?
This is a hack, just like my "select DM_ETH" in our board kconfig now. We want users to be able to configure this via kconfig as well.
Quoted 3 lines
Are you missing a way to call
eth_get_dev()to trigger aforementioned devicetree setup?
Nope, eth_env_set_enetaddr_by_index, and also some networking var flags for setting CONFIG_ENV_FLAGS_LIST_STATIC. That's what my cited patch takes care of.
Quoted 2 lines
Adding 'net list' to
CONFIG_PREBOOTmight do the trick.
We don't enable CONFIG_CMD_NET - because there is no network to configure with it.
Thanks,
Jan
Signature · 2 lines
Siemens AG, Technology Competence Center Embedded Linux
PATCH v2 2/39 efi: Add a separate maintainer entry for the app
In reply to Simon GlassSeparate this out slightly from the payload, with a new entry.
We might consider renaming EFI PAYLOAD to EFI LOADER, but that would require quite a lot of file changes.
---
Changes in v2:
- Add new patch with maintainer entry
MAINTAINERS | 7 +++++++ 1 file changed, 7 insertions(+)
MAINTAINERS +7 -0
@@ -700,6 +700,13 @@ F: drivers/core/ F: include/dm/ F: test/dm/ +EFI APP +M: Simon Glass <[email protected]> +M: Heinrich Schuchardt <[email protected]> +S: Maintained +W: https://u-boot.readthedocs.io/en/latest/develop/uefi/u-boot_on_efi.html +F: lib/efi/efi_app.c + EFI PAYLOAD M: Heinrich Schuchardt <[email protected]> R: Alexander Graf <[email protected]>
Signature · 1 line
2.33.0.685.g46640cef36-goog
Re: [PATCH v2 02/39] efi: Add a separate maintainer entry for the app
Quoted 6 lines · Simon Glass
Separate this out slightly from the payload, with a new entry.
We might consider renaming EFI PAYLOAD to EFI LOADER, but that would require quite a lot of file changes.
Signed-off-by Simon Glass <[email protected]>
Quoted 27 lines
---
Changes in v2:
- Add new patch with maintainer entryMAINTAINERS | 7 +++++++ 1 file changed, 7 insertions(+)MAINTAINERS +7
-0@@ -700,6 +700,13 @@ F: drivers/core/ F: include/dm/ F: test/dm/ +EFI APP +M: Simon Glass <[email protected]> +M: Heinrich Schuchardt <[email protected]> +S: Maintained +W: https://u-boot.readthedocs.io/en/latest/develop/uefi/u-boot_on_efi.html +F: lib/efi/efi_app.c + EFI PAYLOAD M: Heinrich Schuchardt <[email protected]> R: Alexander Graf <[email protected]>
PATCH v2 3/39 x86: Keep symbol information in u-boot ELF file
In reply to Simon GlassAt present this information is stripped when linking. It is useful to keep it around. Strip it from the .efi files instead.
---
(no changes since v1)
arch/x86/config.mk | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
arch/x86/config.mk +2 -2
@@ -37,7 +37,7 @@ KBUILD_LDFLAGS += -m $(if $(IS_32BIT),elf_i386,elf_x86_64) LDFLAGS_EFI_PAYLOAD := -Bsymbolic -Bsymbolic-functions -shared --no-undefined -s OBJCOPYFLAGS_EFI := -j .text -j .sdata -j .data -j .dynamic -j .dynsym \ - -j .rel -j .rela -j .reloc + -j .rel -j .rela -j .reloc --strip-all # Compiler flags to be added when building UEFI applications CFLAGS_EFI := -fpic -fshort-wchar
@@ -65,7 +65,7 @@ CPPFLAGS_crt0-efi-$(EFIARCH).o += $(CFLAGS_EFI) ifeq ($(CONFIG_EFI_APP),y) PLATFORM_CPPFLAGS += $(CFLAGS_EFI) -LDFLAGS_FINAL += -znocombreloc -shared -s +LDFLAGS_FINAL += -znocombreloc -shared LDSCRIPT := $(LDSCRIPT_EFI) else
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 4/39 x86: Create a new header for EFI
In reply to Simon GlassThe setup routines are called from zimage but don't really belong in the zimage header. Add a new EFI header to house these. Add comments so it is clear what the functions do.
Note that these functions are x86-specific. The zimage business is not used on other architectures.
---
Changes in v2:
- Update commit message to mention that these function are x86-specific
arch/x86/include/asm/efi.h | 32 ++++++++++++++++++++++++++++++++ arch/x86/include/asm/zimage.h | 3 --- arch/x86/lib/zimage.c | 1 + 3 files changed, 33 insertions(+), 3 deletions(-) create mode 100644 arch/x86/include/asm/efi.h
arch/x86/include/asm/efi.h new file +32 -0
@@ -0,0 +1,32 @@ +/* SPDX-License-Identifier: GPL-2.0+ */ +/* +* Copyright Google LLC +*/ + +#ifndef _ASM_EFI_H_ +#define _ASM_EFI_H_ + +struct efi_info ; +struct screen_info ; + +/** +* setup_video() - Set up the screen info in the x86 setup +* +* This is needed so Linux can use the display (when U-Boot is an EFI payload) +* +* @efi_info: Pointer to place to put the screen info in the x86 setup base +*/ +void setup_video (struct screen_info *screen_info); + +/** +* setup_efi_info() - Set up the EFI info needed by Linux to boot +* +* This writes a suitable signature, table pointers, memory-map pointer, etc. +* These are needed for Linux to boot from U-Boot (when U-Boot is an EFI +* payload). +* +* @efi_info: Pointer to place to put the EFI info in the x86 setup base +*/ +void setup_efi_info (struct efi_info *efi_info); + +#endif
arch/x86/include/asm/zimage.h +0 -3
@@ -72,7 +72,4 @@ int setup_zimage(struct boot_params *setup_base, char *cmd_line, int auto_boot,*/ void zimage_dump (struct boot_params *base_ptr); -void setup_video (struct screen_info *screen_info); -void setup_efi_info (struct efi_info *efi_info); -#endif
arch/x86/lib/zimage.c +1 -0
@@ -29,6 +29,7 @@#include <asm/byteorder.h> #include <asm/bootm.h> #include <asm/bootparam.h> +#include <asm/efi.h> #include <asm/global_data.h> #ifdef CONFIG_SYS_COREBOOT #include <asm/arch/timestamp.h>
Signature · 1 line
2.33.0.685.g46640cef36-goog
Re: [PATCH v2 04/39] x86: Create a new header for EFI
Quoted 8 lines · Simon Glass
The setup routines are called from zimage but don't really belong in the zimage header. Add a new EFI header to house these. Add comments so it is clear what the functions do.
Note that these functions are x86-specific. The zimage business is not used on other architectures.
Signed-off-by Simon Glass <[email protected]>
Quoted 74 lines
---
Changes in v2:
- Update commit message to mention that these function are x86-specificarch/x86/include/asm/efi.h | 32 ++++++++++++++++++++++++++++++++ arch/x86/include/asm/zimage.h | 3 --- arch/x86/lib/zimage.c | 1 + 3 files changed, 33 insertions(+), 3 deletions(-)create mode 100644
arch/x86/include/asm/efi.harch/x86/include/asm/efi.h new file +32
-0@@ -0,0 +1,32 @@ +/* SPDX-License-Identifier: GPL-2.0+ */ +/* +* Copyright Google LLC +*/ + +#ifndef _ASM_EFI_H_ +#define _ASM_EFI_H_ + +struct efi_info ; +struct screen_info ; + +/** +* setup_video() - Set up the screen info in the x86 setup +* +* This is needed so Linux can use the display (when U-Boot is an EFI payload) +* +* @efi_info: Pointer to place to put the screen info in the x86 setup base +*/ +void setup_video (struct screen_info *screen_info); + +/** +* setup_efi_info() - Set up the EFI info needed by Linux to boot +* +* This writes a suitable signature, table pointers, memory-map pointer, etc. +* These are needed for Linux to boot from U-Boot (when U-Boot is an EFI +* payload). +* +* @efi_info: Pointer to place to put the EFI info in the x86 setup base +*/ +void setup_efi_info (struct efi_info *efi_info); + +#endif arch/x86/include/asm/zimage.h +0
-3@@ -72,7 +72,4 @@ int setup_zimage(struct boot_params *setup_base, char *cmd_line, int auto_boot,*/ void zimage_dump (struct boot_params *base_ptr); -void setup_video (struct screen_info *screen_info); -void setup_efi_info (struct efi_info *efi_info); -#endif arch/x86/lib/zimage.c +1
-0@@ -29,6 +29,7 @@#include <asm/byteorder.h> #include <asm/bootm.h> #include <asm/bootparam.h> +#include <asm/efi.h> #include <asm/global_data.h> #ifdef CONFIG_SYS_COREBOOT #include <asm/arch/timestamp.h>
PATCH v2 5/39 x86: Show some EFI info with the bdinfo command
In reply to Simon GlassIt is useful to see some basic EFI info with the command as it forms part of the information about a board.
Add a hook for this and show the table address as a start.
While here, fix an invalid cast in setup_efi_info(). Note that this function is using a data structure defined by Linux so we cannot change it. Also note that ulong is used since this is the standard in U-Boot (>6k uses), despite there being quite a bit of the more verbose uintptr_t (930 uses).
---
Changes in v2:
- Update commit message to mention why the data structure is like it is
- Update commit message to mention why ulong is used
arch/x86/cpu/efi/payload.c | 13 +++++++++++-- arch/x86/include/asm/efi.h | 7 +++++++ arch/x86/lib/Makefile | 1 + arch/x86/lib/bdinfo.c | 22 ++++++++++++++++++++++ 4 files changed, 41 insertions(+), 2 deletions(-) create mode 100644 arch/x86/lib/bdinfo.c
arch/x86/cpu/efi/payload.c +11 -2
@@ -280,15 +280,24 @@ void setup_efi_info(struct efi_info *efi_info) } efi_info->efi_memdesc_size = map->desc_size; efi_info->efi_memdesc_version = map->version; - efi_info->efi_memmap = (u32)(map->desc); + efi_info->efi_memmap = (ulong)(map->desc); efi_info->efi_memmap_size = size -sizeof (struct efi_entry_memmap );#ifdef CONFIG_EFI_STUB_64BIT efi_info->efi_systab_hi = table->sys_table >>32 ; - efi_info->efi_memmap_hi = (u64)(u32)(map->desc) >>32 ; + efi_info->efi_memmap_hi = (u64 )(ulong )map->desc >>32 ; signature =EFI64_LOADER_SIGNATURE ;#else signature =EFI32_LOADER_SIGNATURE ;#endif memcpy (&efi_info->efi_loader_signature, signature,4 ); } + +void efi_show_bdinfo (void ) +{ +struct efi_entry_systable *table =NULL ; +int size, ret; + + ret =efi_info_get (EFIET_SYS_TABLE , (void **)&table, &size); +bdinfo_print_num_l ("efi_table" , (ulong )table); +}
arch/x86/include/asm/efi.h +7 -0
@@ -29,4 +29,11 @@ void setup_video(struct screen_info *screen_info);*/ void setup_efi_info (struct efi_info *efi_info); +/** +* efi_show_bdinfo() - Show information about EFI for the 'bdinfo' command +* +* This looks up the EFI table pointer and shows related info +*/ +void efi_show_bdinfo (void ); +#endif
arch/x86/lib/Makefile +1 -0
@@ -3,6 +3,7 @@ # (C) Copyright 2002-2006 # Wolfgang Denk, DENX Software Engineering, [email protected]. +obj-y += bdinfo.o ifndef CONFIG_X86_64 ifndef CONFIG_TPL_BUILD obj-y += bios.o
arch/x86/lib/bdinfo.c new file +22 -0
@@ -0,0 +1,22 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* x86-specific information for the 'bd' command +* +* Copyright 2021 Google LLC +*/ + +#include <common.h> +#include <efi.h> +#include <init.h> +#include <asm/efi.h> +#include <asm/global_data.h> + +DECLARE_GLOBAL_DATA_PTR ; + +void arch_print_bdinfo (void ) +{ +bdinfo_print_num_l ("prev table" , gd->arch.table); + +if (IS_ENABLED (CONFIG_EFI_STUB )) +efi_show_bdinfo (); +}
Signature · 1 line
2.33.0.685.g46640cef36-goog
Re: [PATCH v2 05/39] x86: Show some EFI info with the bdinfo command
Quoted 45 lines · Simon Glass
It is useful to see some basic EFI info with the command as it forms part of the information about a board.
Add a hook for this and show the table address as a start.
While here, fix an invalid cast in
setup_efi_info(). Note that this function is using a data structure defined by Linux so we cannot change it. Also note that ulong is used since this is the standard in U-Boot (>6k uses), despite there being quite a bit of the more verboseuintptr_t(930 uses).Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- Update commit message to mention why the data structure is like it is
- Update commit message to mention why ulong is usedarch/x86/cpu/efi/payload.c | 13 +++++++++++-- arch/x86/include/asm/efi.h | 7 +++++++ arch/x86/lib/Makefile | 1 + arch/x86/lib/bdinfo.c | 22 ++++++++++++++++++++++ 4 files changed, 41 insertions(+), 2 deletions(-)create mode 100644
arch/x86/lib/bdinfo.carch/x86/cpu/efi/payload.c +2
-2@@ -280,15 +280,24 @@ void setup_efi_info(struct efi_info *efi_info) } efi_info->efi_memdesc_size = map->desc_size; efi_info->efi_memdesc_version = map->version; - efi_info->efi_memmap = (u32)(map->desc); + efi_info->efi_memmap = (ulong)(map->desc); efi_info->efi_memmap_size = size -sizeof (struct efi_entry_memmap );#ifdef CONFIG_EFI_STUB_64BIT efi_info->efi_systab_hi = table->sys_table >>32 ; - efi_info->efi_memmap_hi = (u64)(u32)(map->desc) >>32 ; + efi_info->efi_memmap_hi = (u64 )(ulong )map->desc >>32 ; signature =EFI64_LOADER_SIGNATURE ;#else signature =EFI32_LOADER_SIGNATURE ;#endif
arch/x86/cpu/efi/payload.c void setup_efi_info(struct efi_info *efi_info) · line 302
Quoted 10 lines
arch/x86/cpu/efi/payload.c +8
-0memcpy (&efi_info->efi_loader_signature, signature,4 ); } + +void efi_show_bdinfo (void ) +{ +struct efi_entry_systable *table =NULL ; +int size, ret; + + ret =efi_info_get (EFIET_SYS_TABLE , (void **)&table, &size); +bdinfo_print_num_l ("efi_table" , (ulong )table);
Some properties of the UEFI API depend on the UEFI API version (EFI_SYSTEM_TABLE.FirmwareRevision)
The revision of the EFI_FILE_PROTOCOL may also be relevant for U-Boot.
Maybe you want to add this information in a later patch.
Quoted 58 lines
arch/x86/cpu/efi/payload.c +1
-0+}arch/x86/include/asm/efi.h +7
-0@@ -29,4 +29,11 @@ void setup_video(struct screen_info *screen_info);*/ void setup_efi_info (struct efi_info *efi_info); +/** +* efi_show_bdinfo() - Show information about EFI for the 'bdinfo' command +* +* This looks up the EFI table pointer and shows related info +*/ +void efi_show_bdinfo (void ); +#endif arch/x86/lib/Makefile +1
-0@@ -3,6 +3,7 @@ # (C) Copyright 2002-2006 # Wolfgang Denk, DENX Software Engineering, [email protected]. +obj-y += bdinfo.o ifndef CONFIG_X86_64 ifndef CONFIG_TPL_BUILD obj-y += bios.oarch/x86/lib/bdinfo.c new file +22
-0@@ -0,0 +1,22 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* x86-specific information for the 'bd' command +* +* Copyright 2021 Google LLC +*/ + +#include <common.h> +#include <efi.h> +#include <init.h> +#include <asm/efi.h> +#include <asm/global_data.h> + +DECLARE_GLOBAL_DATA_PTR ; + +void arch_print_bdinfo (void ) +{ +bdinfo_print_num_l ("prev table" , gd->arch.table); + +if (IS_ENABLED (CONFIG_EFI_STUB )) +efi_show_bdinfo (); +}
Re: [PATCH v2 05/39] x86: Show some EFI info with the bdinfo command
Hi Heinrich,
Quoted 52 lines · Heinrich Schuchardt
Quoted 50 lines · Simon Glass
It is useful to see some basic EFI info with the command as it forms part of the information about a board.
Add a hook for this and show the table address as a start.
While here, fix an invalid cast in
setup_efi_info(). Note that this function is using a data structure defined by Linux so we cannot change it. Also note that ulong is used since this is the standard in U-Boot (>6k uses), despite there being quite a bit of the more verboseuintptr_t(930 uses).Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- Update commit message to mention why the data structure is like it is
- Update commit message to mention why ulong is usedarch/x86/cpu/efi/payload.c | 13 +++++++++++-- arch/x86/include/asm/efi.h | 7 +++++++ arch/x86/lib/Makefile | 1 + arch/x86/lib/bdinfo.c | 22 ++++++++++++++++++++++ 4 files changed, 41 insertions(+), 2 deletions(-)create mode 100644
arch/x86/lib/bdinfo.carch/x86/cpu/efi/payload.c +5
-2@@ -280,15 +280,24 @@ void setup_efi_info(struct efi_info *efi_info) } efi_info->efi_memdesc_size = map->desc_size; efi_info->efi_memdesc_version = map->version; - efi_info->efi_memmap = (u32)(map->desc); + efi_info->efi_memmap = (ulong)(map->desc); efi_info->efi_memmap_size = size -sizeof (struct efi_entry_memmap );#ifdef CONFIG_EFI_STUB_64BIT efi_info->efi_systab_hi = table->sys_table >>32 ; - efi_info->efi_memmap_hi = (u64)(u32)(map->desc) >>32 ; + efi_info->efi_memmap_hi = (u64 )(ulong )map->desc >>32 ; signature =EFI64_LOADER_SIGNATURE ;#else signature =EFI32_LOADER_SIGNATURE ;#endif memcpy (&efi_info->efi_loader_signature, signature,4 ); } + +void efi_show_bdinfo (void ) +{
Quoted 8 lines
arch/x86/cpu/efi/payload.c void setup_efi_info(struct efi_info *efi_info) · line 302
Quoted 5 lines
arch/x86/cpu/efi/payload.c +5
-0+struct efi_entry_systable *table =NULL ; +int size, ret; + + ret =efi_info_get (EFIET_SYS_TABLE , (void **)&table, &size); +bdinfo_print_num_l ("efi_table" , (ulong )table);Some properties of the UEFI API depend on the UEFI API version (EFI_SYSTEM_TABLE.FirmwareRevision)
OK I will add a patch.
Quoted 4 lines
The revision of the
EFI_FILE_PROTOCOLmay also be relevant for U-Boot.Maybe you want to add this information in a later patch.
This is tricky because the protocol version is no-longer available once the payload has loaded.
[..]
Regards,
Simon
PATCH v2 6/39 x86: Tidy up global_data pointer for 64-bit
In reply to Simon GlassAdd an extern declaration so that it is possible to use this macro in files other than the one that defines it.
---
(no changes since v1)
arch/x86/cpu/x86_64/cpu.c | 3 +++ arch/x86/include/asm/global_data.h | 2 ++ 2 files changed, 5 insertions(+)
arch/x86/cpu/x86_64/cpu.c +3 -0
@@ -8,6 +8,9 @@#include <cpu_func.h> #include <debug_uart.h> #include <init.h> +#include <asm/global_data.h> + +DECLARE_GLOBAL_DATA_PTR ;/* * Global declaration of gd.
arch/x86/include/asm/global_data.h +2 -0
@@ -133,6 +133,8 @@ struct arch_global_data {#ifndef __ASSEMBLY__# if defined(CONFIG_EFI_APP ) ||CONFIG_IS_ENABLED (X86_64 ) +extern struct global_data *global_data_ptr; +/* TODO([email protected]): Consider using a fixed register for gd on x86_64 */ #define gd global_data_ptr
Signature · 1 line
2.33.0.685.g46640cef36-goog
Re: [PATCH v2 06/39] x86: Tidy up global_data pointer for 64-bit
Quoted 32 lines · Simon Glass
Add an extern declaration so that it is possible to use this macro in files other than the one that defines it.
Signed-off-by Simon Glass <[email protected]>Reviewed-by Heinrich Schuchardt <[email protected]>---
(no changes since v1)
arch/x86/cpu/x86_64/cpu.c | 3 +++ arch/x86/include/asm/global_data.h | 2 ++ 2 files changed, 5 insertions(+)arch/x86/cpu/x86_64/cpu.c +3
-0@@ -8,6 +8,9 @@#include <cpu_func.h> #include <debug_uart.h> #include <init.h> +#include <asm/global_data.h> + +DECLARE_GLOBAL_DATA_PTR ;/* * Global declaration of gd. arch/x86/include/asm/global_data.h +0
-0
arch/x86/include/asm/global_data.h struct arch_global_data { · line 136
Quoted 4 lines
arch/x86/include/asm/global_data.h +1
-0@@ -133,6 +133,8 @@ struct arch_global_data {#ifndef __ASSEMBLY__# if defined(CONFIG_EFI_APP ) ||CONFIG_IS_ENABLED (X86_64 ) +extern struct global_data *global_data_ptr;
You already have:arch/x86/include/asm/global_data.h:139:
#define DECLARE_GLOBAL_DATA_PTR extern struct global_data *global_data_ptr
Why don't you use DECLARE_GLOBAL_DATA_PTR where needed?
Best regards
Heinrich
Quoted 5 lines
arch/x86/include/asm/global_data.h +1
-0+/* TODO([email protected]): Consider using a fixed register for gd on x86_64 */ #define gd global_data_ptr
Re: [PATCH v2 06/39] x86: Tidy up global_data pointer for 64-bit
Hi Heinrich,
Quoted 37 lines · Heinrich Schuchardt
Quoted 35 lines · Simon Glass
Add an extern declaration so that it is possible to use this macro in files other than the one that defines it.
Signed-off-by Simon Glass <[email protected]>Reviewed-by Heinrich Schuchardt <[email protected]>---
(no changes since v1)
arch/x86/cpu/x86_64/cpu.c | 3 +++ arch/x86/include/asm/global_data.h | 2 ++ 2 files changed, 5 insertions(+)arch/x86/cpu/x86_64/cpu.c +3
-0@@ -8,6 +8,9 @@#include <cpu_func.h> #include <debug_uart.h> #include <init.h> +#include <asm/global_data.h> + +DECLARE_GLOBAL_DATA_PTR ;/* * Global declaration of gd. arch/x86/include/asm/global_data.h +0
-0@@ -133,6 +133,8 @@ struct arch_global_data {#ifndef __ASSEMBLY__# if defined(CONFIG_EFI_APP ) ||CONFIG_IS_ENABLED (X86_64 )
Quoted 7 lines
arch/x86/include/asm/global_data.h struct arch_global_data { · line 136
Quoted 1 line
arch/x86/include/asm/global_data.h +1
-0+extern struct global_data *global_data_ptr;You already have:
arch/x86/include/asm/global_data.h:139:
#defineDECLARE_GLOBAL_DATA_PTRextern structglobal_data*global_data_ptrWhy don't you use
DECLARE_GLOBAL_DATA_PTRwhere needed?
I have no idea what I was thinking or how I got into this state. But anyway, I don't think we need this patch. Thanks for catching it.
Regards,
Simon
PATCH v2 7/39 efi: Add a script for building and testing U-Boot on UEFI
In reply to Simon GlassIt is quite complicating to run U-Boot on qemu since we have four different builds and they must use different versions of qemu and the UEFI binaries.
Add a script to help.
---
Changes in v2:
- Add MAINTAINERS entry
- Add support for creating a partition table with a filesystem inside
- Add support for running qemu with just a serial console (no display)
MAINTAINERS | 1 + doc/develop/uefi/u-boot_on_efi.rst | 62 ++++++++++ scripts/build-efi.sh | 188 +++++++++++++++++++++++++++++ 3 files changed, 251 insertions(+) create mode 100755 scripts/build-efi.sh
MAINTAINERS +1 -0
@@ -706,6 +706,7 @@ M: Heinrich Schuchardt <[email protected]> S: Maintained W: https://u-boot.readthedocs.io/en/latest/develop/uefi/u-boot_on_efi.html F: lib/efi/efi_app.c +F: scripts/build-efi.sh EFI PAYLOAD M: Heinrich Schuchardt <[email protected]>
doc/develop/uefi/u-boot_on_efi.rst +62 -0
74 lines not on this page · raw message
scripts/build-efi.sh new file +188 -0
188 lines not on this page · raw message
Signature · 1 line
2.33.0.685.g46640cef36-goog
Re: [PATCH v2 07/39] efi: Add a script for building and testing U-Boot on UEFI
Quoted 1 line · Simon Glass
It is quite complicating to run U-Boot on qemu since we have four
Nits:
%s/complicating/complicated/ %/qemu/QEMU/
Quoted 1 line
different builds and they must use different versions of qemu and the
QEMU
Quoted 133 lines
UEFI binaries.
Add a script to help.
Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- AddMAINTAINERSentry
- Add support for creating a partition table with a filesystem inside
- Add support for running qemu with just a serial console (no display)MAINTAINERS | 1 + doc/develop/uefi/u-boot_on_efi.rst | 62 ++++++++++ scripts/build-efi.sh | 188 +++++++++++++++++++++++++++++ 3 files changed, 251 insertions(+)create mode 100755
scripts/build-efi.shMAINTAINERS +1
-0@@ -706,6 +706,7 @@ M: Heinrich Schuchardt <[email protected]> S: Maintained W: https://u-boot.readthedocs.io/en/latest/develop/uefi/u-boot_on_efi.html F: lib/efi/efi_app.c +F: scripts/build-efi.sh EFI PAYLOAD M: Heinrich Schuchardt <[email protected]>doc/develop/uefi/u-boot_on_efi.rst +62
-0@@ -96,6 +96,11 @@ that EFI does not support booting a 64-bit application from a 32-bit EFI (or vice versa). Also it will often fail to print an error message if you get this wrong. +You may find the script `scripts/build-efi.sh` helpful for building and testing +U-Boot on UEFI on QEMU. It also includes links to UEFI binaries dating from +2021. + +See `Example run`_ for an example run. Inner workings --------------@@ -191,6 +196,63 @@ of code is built this way (see the extra- line in lib/efi/Makefile). Everything else is built as a normal U-Boot, so is always 32-bit on x86 at present. +Example run +----------- + +This shows running with serial enabled (see `include/configs/efi-x86_app.h`):: + + $ scripts/build-efi.sh -wsPr + Packaging efi-x86_app32 + Running qemu-system-i386 + + BdsDxe: failed to load Boot0001 "UEFI QEMU HARDDISK QM00005 " from PciRoot(0x0)/Pci(0x3,0x0)/Sata(0x0,0xFFFF,0x0): Not Found + BdsDxe: loading Boot0002 "EFI Internal Shell" from Fv(7CB8BDC9-F8EB-4F34-AAEA-3EE4AF6516A1)/FvFile(7C04A583-9E3E-4F1C-AD65-E05268D0B4D1) + BdsDxe: starting Boot0002 "EFI Internal Shell" from Fv(7CB8BDC9-F8EB-4F34-AAEA-3EE4AF6516A1)/FvFile(7C04A583-9E3E-4F1C-AD65-E05268D0B4D1) + + UEFI Interactive Shell v2.2 + EDK II + UEFI v2.70 (EDK II, 0x00010000) + Mapping table + FS0: Alias(s):HD0a65535a1:;BLK1: + PciRoot(0x0)/Pci(0x3,0x0)/Sata(0x0,0xFFFF,0x0)/HD(1,GPT,0FFD5E61-3B0C-4326-8049-BDCDC910AF72,0x800,0xB000) + BLK0: Alias(s): + PciRoot(0x0)/Pci(0x3,0x0)/Sata(0x0,0xFFFF,0x0) + + Press ESC in 5 seconds to skip startup.nsh or any other key to continue. + Shell> fs0:u-boot-app.efi + U-Boot EFI App (using allocated RAM address 47d4000) key=8d4, image=06a6f610 + starting + + + U-Boot 2022.01-rc4 (Sep 19 2021 - 14:03:20 -0600) + + CPU: x86, vendor Intel, device 663h + DRAM: 32 MiB + 0: efi_media_0 PciRoot(0x0)/Pci(0x3,0x0)/Sata(0x0,0xFFFF,0x0) + 1: <partition> PciRoot(0x0)/Pci(0x3,0x0)/Sata(0x0,0xFFFF,0x0)/HD(1,GPT,0FFD5E61-3B0C-4326-8049-BDCDC910AF72,0x800,0xB000) + Loading Environment from nowhere... OK + Model: EFI x86 Application + Hit any key to stop autoboot: 0 + + Partition Map for EFI device 0 -- Partition Type: EFI + + Part Start LBA End LBA Name + Attributes + Type GUID + Partition GUID + 1 0x00000800 0x0000b7ff "boot" + attrs: 0x0000000000000000 + type: ebd0a0a2-b9e5-4433-87c0-68b6b72699c7 + guid: 0ffd5e61-3b0c-4326-8049-bdcdc910af72 + 19 startup.nsh + 528384 u-boot-app.efi + 10181 NvVars + + 3 file(s), 0 dir(s) + + => QEMU: Terminated + + Future work ----------- This work could be extended in a number of ways:scripts/build-efi.sh new file +17
-0@@ -0,0 +1,188 @@ +#!/bin/bash +# SPDX-License-Identifier: GPL-2.0+ +# +# Script to build an EFI thing suitable for booting with qemu, possibly running +# it also. + +# This just an example. It assumes that + +# - you build U-Boot in /tmp/b/<name> where <name> is the U-Boot board config +# - /mnt/x is a directory used for mounting +# - you have access to the 'pure UEFI' builds for qemu +# +# UEFI binaries for QEMU used for testing this script: +# +# OVMF-pure-efi.i386.fd at +# https://drive.google.com/file/d/1jWzOAZfQqMmS2_dAK2G518GhIgj9r2RY/view?usp=sharing +
scripts/build-efi.sh line 27
Quoted 10 lines
scripts/build-efi.sh new file +10
-0+# OVMF-pure-efi.x64.fd at +# https://drive.google.com/file/d/1c39YI9QtpByGQ4V0UNNQtGqttEzS-eFV/view?usp=sharing + +set -e + +usage() { + echo "Usage: $0 [-a | -p] [other opts]" 1>&2 + echo 1>&2 + echo " -a - Package up the app" 1>&2 + echo " -o - Use old EFI app build (before 32/64 split)" 1>&2
Why is this option needed?
scripts/build-efi.sh line 30
Quoted 3 lines
scripts/build-efi.sh new file +3
-0+ echo " -p - Package up the payload" 1>&2 + echo " -P - Create a partition table" 1>&2 + echo " -r - Run qemu with the image" 1>&2
QEMU
scripts/build-efi.sh line 31
Quoted 1 line
scripts/build-efi.sh new file +1
-0+ echo " -s - Run qemu with serial only (no display)" 1>&2
QEMU
Best regards
Heinrich
Quoted 158 lines
scripts/build-efi.sh new file +157
-0+ echo " -w - Use word version (32-bit)" 1>&2 + exit 1 +} + +# 32- or 64-bit EFI +bitness=64 + +# app or payload ? +type=app + +# create a partition table and put the filesystem in that (otherwise put the +# filesystem in the raw device) +part= + +# run the image with qemu +run= + +# run qemu without a display (U-Boot must be set to stdout=serial) +serial= + +# before the 32/64 split of the app +old= + +while getopts "aopPrsw" opt; do + case "${opt}" in + a) + type=app + ;; + p) + type=payload + ;; + r) + run=1 + ;; + s) + serial=1 + ;; + w) + bitness=32 + ;; + o) + old=1 + ;; + P) + part=1 + ;; + *) + usage + ;; + esac +done + +run_qemu() { + extra= + if [[ "${bitness}" = "64" ]]; then + qemu=qemu-system-x86_64 + bios=OVMF-pure-efi.x64.fd + else + qemu=qemu-system-i386 + bios=OVMF-pure-efi.i386.fd + fi + if [[ -n "${serial}" ]]; then + extra="-display none -serial mon:stdio" + fi + echo "Running ${qemu}" + "${qemu}" -bios "${bios}" \ + -drive id=disk,file="${IMG}",if=none,format=raw \ + -nic none -device ahci,id=ahci \ + -device ide-hd,drive=disk,bus=ahci.0 ${extra} +} + +# Put files in /tmp/b/${BUILD} +setup_files() { + echo "Packaging ${BUILD}" + mkdir -p $TMP + cat >$TMP/startup.nsh <<EOF +fs0:u-boot-${type}.efi +EOF + sudo cp /tmp/b/$BUILD/u-boot-${type}.efi $TMP + + # Can copy in other files here: + #sudo cp /tmp/b/$BUILD/image.bin $TMP/chromeos.rom + #sudo cp /boot/vmlinuz-5.4.0-77-generic $TMP/vmlinuz +} + +# Copy files into the filesystem +copy_files() { + sudo cp $TMP/* $MNT +} + +# Create a filesystem on a raw device and copy in the files +setup_raw() { + mkfs.vfat "${IMG}" >/dev/null + sudo mount -o loop "${IMG}" $MNT + copy_files + sudo umount $MNT +} + +# Create a partition table and put the filesystem in the first partition +# then copy in the files +setup_part() { + # Create a gpt partition table with one parittion + parted "${IMG}" mklabel gpt 2>/dev/null + + # This doesn't work correctly. It creates: + # Number Start End Size File system Name Flags + # 1 1049kB 24.1MB 23.1MB boot msftdata + # Odd if the same is entered interactively it does set the FS type + parted -s -a optimal -- "${IMG}" mkpart boot fat32 1MiB 23MiB + + # Map this partition to a loop device + kp="$(sudo kpartx -av ${IMG})" + read boot_dev<<<$(grep -o 'loop.*p.' <<< "${kp}") + test "${boot_dev}" + dev="/dev/mapper/${boot_dev}" + + mkfs.vfat "${dev}" >/dev/null + + sudo mount -o loop "${dev}" $MNT + + copy_files + + # Sync here since this makes kpartx more likely to work the first time + sync + sudo umount $MNT + + # For some reason this needs a sleep or it sometimes fails, if it was + # run recently (in the last few seconds) + if ! sudo kpartx -d "${IMG}" > /dev/null; then + sleep .5 + sudo kpartx -d "${IMG}" > /dev/null || \ + echo "Failed to remove ${boot_dev}, use: sudo kpartx -d ${IMG}" + fi +} + +TMP="/tmp/efi${bitness}${type}" +MNT=/mnt/x +BUILD="efi-x86_${type}${bitness}" +IMG=try.img + +if [[ -n "${old}" && "${bitness}" = "32" ]]; then + BUILD="efi-x86_${type}" +fi + +setup_files + +qemu-img create "${IMG}" 24M >/dev/null + +if [[ -n "${part}" ]]; then + setup_part +else + setup_raw +fi + +if [[ -n "${run}" ]]; then + run_qemu +fi
Re: [PATCH v2 07/39] efi: Add a script for building and testing U-Boot on UEFI
Hi Heinrich,
Quoted 28 lines · Heinrich Schuchardt
Quoted 1 line · Simon Glass
It is quite complicating to run U-Boot on qemu since we have four
Nits:
%s/complicating/complicated/ %/qemu/QEMU/Quoted 1 line
different builds and they must use different versions of qemu and the
QEMU
Quoted 17 lines
UEFI binaries.
Add a script to help.
Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- AddMAINTAINERSentry
- Add support for creating a partition table with a filesystem inside
- Add support for running qemu with just a serial console (no display)MAINTAINERS | 1 + doc/develop/uefi/u-boot_on_efi.rst | 62 ++++++++++ scripts/build-efi.sh | 188 +++++++++++++++++++++++++++++ 3 files changed, 251 insertions(+)create mode 100755
scripts/build-efi.sh
[..]
Quoted 27 lines
Quoted 27 lines
scripts/build-efi.sh new file +21
-0@@ -0,0 +1,188 @@ +#!/bin/bash +# SPDX-License-Identifier: GPL-2.0+ +# +# Script to build an EFI thing suitable for booting with qemu, possibly running +# it also. + +# This just an example. It assumes that + +# - you build U-Boot in /tmp/b/<name> where <name> is the U-Boot board config +# - /mnt/x is a directory used for mounting +# - you have access to the 'pure UEFI' builds for qemu +# +# UEFI binaries for QEMU used for testing this script: +# +# OVMF-pure-efi.i386.fd at +# https://drive.google.com/file/d/1jWzOAZfQqMmS2_dAK2G518GhIgj9r2RY/view?usp=sharing + +# OVMF-pure-efi.x64.fd at +# https://drive.google.com/file/d/1c39YI9QtpByGQ4V0UNNQtGqttEzS-eFV/view?usp=sharing + +set -e
Quoted 8 lines
scripts/build-efi.sh line 27
Quoted 6 lines
scripts/build-efi.sh new file +6
-0+ +usage() { + echo "Usage: $0 [-a | -p] [other opts]" 1>&2 + echo 1>&2 + echo " -a - Package up the app" 1>&2 + echo " -o - Use old EFI app build (before 32/64 split)" 1>&2Why is this option needed?
I think it is wise for now, in case we hit regressions. This allows comparing the 'old' build with the new one. I have used it quite a few times. We can drop this option once everything is settled.
[..]
Regards,
Simon
Re: [PATCH v2 07/39] efi: Add a script for building and testing U-Boot on UEFI
In reply to Simon GlassQuoted 7 lines · Simon Glass
It is quite complicating to run U-Boot on qemu since we have four different builds and they must use different versions of qemu and the UEFI binaries.
Add a script to help.
Signed-off-by Simon Glass <[email protected]>
This patch seems not to work:
$ scripts/build-efi.sh -wsPr Packaging efi-x86_app32 cp: cannot stat '/tmp/b/efi-x86_app32/u-boot-app.efi': No such file or director
The scripts lacks a step to build the EFI binary.
If you expect the user to build it, you will find it in the current directory and not in some arbitrary subdirectory of /tmp.
Temporary directories should be created with 'mktemp -d'. Using fixed paths (e.g. /tmp/b/) may lead to conflicts.
Best regards
Heinrich
Quoted 300 lines
---
Changes in v2:
- AddMAINTAINERSentry
- Add support for creating a partition table with a filesystem inside
- Add support for running qemu with just a serial console (no display)MAINTAINERS | 1 + doc/develop/uefi/u-boot_on_efi.rst | 62 ++++++++++ scripts/build-efi.sh | 188 +++++++++++++++++++++++++++++ 3 files changed, 251 insertions(+)create mode 100755
scripts/build-efi.shMAINTAINERS +1
-0@@ -706,6 +706,7 @@ M: Heinrich Schuchardt <[email protected]> S: Maintained W: https://u-boot.readthedocs.io/en/latest/develop/uefi/u-boot_on_efi.html F: lib/efi/efi_app.c +F: scripts/build-efi.sh EFI PAYLOAD M: Heinrich Schuchardt <[email protected]>doc/develop/uefi/u-boot_on_efi.rst +62
-0@@ -96,6 +96,11 @@ that EFI does not support booting a 64-bit application from a 32-bit EFI (or vice versa). Also it will often fail to print an error message if you get this wrong. +You may find the script `scripts/build-efi.sh` helpful for building and testing +U-Boot on UEFI on QEMU. It also includes links to UEFI binaries dating from +2021. + +See `Example run`_ for an example run. Inner workings --------------@@ -191,6 +196,63 @@ of code is built this way (see the extra- line in lib/efi/Makefile). Everything else is built as a normal U-Boot, so is always 32-bit on x86 at present. +Example run +----------- + +This shows running with serial enabled (see `include/configs/efi-x86_app.h`):: + + $ scripts/build-efi.sh -wsPr + Packaging efi-x86_app32 + Running qemu-system-i386 + + BdsDxe: failed to load Boot0001 "UEFI QEMU HARDDISK QM00005 " from PciRoot(0x0)/Pci(0x3,0x0)/Sata(0x0,0xFFFF,0x0): Not Found + BdsDxe: loading Boot0002 "EFI Internal Shell" from Fv(7CB8BDC9-F8EB-4F34-AAEA-3EE4AF6516A1)/FvFile(7C04A583-9E3E-4F1C-AD65-E05268D0B4D1) + BdsDxe: starting Boot0002 "EFI Internal Shell" from Fv(7CB8BDC9-F8EB-4F34-AAEA-3EE4AF6516A1)/FvFile(7C04A583-9E3E-4F1C-AD65-E05268D0B4D1) + + UEFI Interactive Shell v2.2 + EDK II + UEFI v2.70 (EDK II, 0x00010000) + Mapping table + FS0: Alias(s):HD0a65535a1:;BLK1: + PciRoot(0x0)/Pci(0x3,0x0)/Sata(0x0,0xFFFF,0x0)/HD(1,GPT,0FFD5E61-3B0C-4326-8049-BDCDC910AF72,0x800,0xB000) + BLK0: Alias(s): + PciRoot(0x0)/Pci(0x3,0x0)/Sata(0x0,0xFFFF,0x0) + + Press ESC in 5 seconds to skip startup.nsh or any other key to continue. + Shell> fs0:u-boot-app.efi + U-Boot EFI App (using allocated RAM address 47d4000) key=8d4, image=06a6f610 + starting + + + U-Boot 2022.01-rc4 (Sep 19 2021 - 14:03:20 -0600) + + CPU: x86, vendor Intel, device 663h + DRAM: 32 MiB + 0: efi_media_0 PciRoot(0x0)/Pci(0x3,0x0)/Sata(0x0,0xFFFF,0x0) + 1: <partition> PciRoot(0x0)/Pci(0x3,0x0)/Sata(0x0,0xFFFF,0x0)/HD(1,GPT,0FFD5E61-3B0C-4326-8049-BDCDC910AF72,0x800,0xB000) + Loading Environment from nowhere... OK + Model: EFI x86 Application + Hit any key to stop autoboot: 0 + + Partition Map for EFI device 0 -- Partition Type: EFI + + Part Start LBA End LBA Name + Attributes + Type GUID + Partition GUID + 1 0x00000800 0x0000b7ff "boot" + attrs: 0x0000000000000000 + type: ebd0a0a2-b9e5-4433-87c0-68b6b72699c7 + guid: 0ffd5e61-3b0c-4326-8049-bdcdc910af72 + 19 startup.nsh + 528384 u-boot-app.efi + 10181 NvVars + + 3 file(s), 0 dir(s) + + => QEMU: Terminated + + Future work ----------- This work could be extended in a number of ways:scripts/build-efi.sh new file +188
-0@@ -0,0 +1,188 @@ +#!/bin/bash +# SPDX-License-Identifier: GPL-2.0+ +# +# Script to build an EFI thing suitable for booting with qemu, possibly running +# it also. + +# This just an example. It assumes that + +# - you build U-Boot in /tmp/b/<name> where <name> is the U-Boot board config +# - /mnt/x is a directory used for mounting +# - you have access to the 'pure UEFI' builds for qemu +# +# UEFI binaries for QEMU used for testing this script: +# +# OVMF-pure-efi.i386.fd at +# https://drive.google.com/file/d/1jWzOAZfQqMmS2_dAK2G518GhIgj9r2RY/view?usp=sharing + +# OVMF-pure-efi.x64.fd at +# https://drive.google.com/file/d/1c39YI9QtpByGQ4V0UNNQtGqttEzS-eFV/view?usp=sharing + +set -e + +usage() { + echo "Usage: $0 [-a | -p] [other opts]" 1>&2 + echo 1>&2 + echo " -a - Package up the app" 1>&2 + echo " -o - Use old EFI app build (before 32/64 split)" 1>&2 + echo " -p - Package up the payload" 1>&2 + echo " -P - Create a partition table" 1>&2 + echo " -r - Run qemu with the image" 1>&2 + echo " -s - Run qemu with serial only (no display)" 1>&2 + echo " -w - Use word version (32-bit)" 1>&2 + exit 1 +} + +# 32- or 64-bit EFI +bitness=64 + +# app or payload ? +type=app + +# create a partition table and put the filesystem in that (otherwise put the +# filesystem in the raw device) +part= + +# run the image with qemu +run= + +# run qemu without a display (U-Boot must be set to stdout=serial) +serial= + +# before the 32/64 split of the app +old= + +while getopts "aopPrsw" opt; do + case "${opt}" in + a) + type=app + ;; + p) + type=payload + ;; + r) + run=1 + ;; + s) + serial=1 + ;; + w) + bitness=32 + ;; + o) + old=1 + ;; + P) + part=1 + ;; + *) + usage + ;; + esac +done + +run_qemu() { + extra= + if [[ "${bitness}" = "64" ]]; then + qemu=qemu-system-x86_64 + bios=OVMF-pure-efi.x64.fd + else + qemu=qemu-system-i386 + bios=OVMF-pure-efi.i386.fd + fi + if [[ -n "${serial}" ]]; then + extra="-display none -serial mon:stdio" + fi + echo "Running ${qemu}" + "${qemu}" -bios "${bios}" \ + -drive id=disk,file="${IMG}",if=none,format=raw \ + -nic none -device ahci,id=ahci \ + -device ide-hd,drive=disk,bus=ahci.0 ${extra} +} + +# Put files in /tmp/b/${BUILD} +setup_files() { + echo "Packaging ${BUILD}" + mkdir -p $TMP + cat >$TMP/startup.nsh <<EOF +fs0:u-boot-${type}.efi +EOF + sudo cp /tmp/b/$BUILD/u-boot-${type}.efi $TMP + + # Can copy in other files here: + #sudo cp /tmp/b/$BUILD/image.bin $TMP/chromeos.rom + #sudo cp /boot/vmlinuz-5.4.0-77-generic $TMP/vmlinuz +} + +# Copy files into the filesystem +copy_files() { + sudo cp $TMP/* $MNT +} + +# Create a filesystem on a raw device and copy in the files +setup_raw() { + mkfs.vfat "${IMG}" >/dev/null + sudo mount -o loop "${IMG}" $MNT + copy_files + sudo umount $MNT +} + +# Create a partition table and put the filesystem in the first partition +# then copy in the files +setup_part() { + # Create a gpt partition table with one parittion + parted "${IMG}" mklabel gpt 2>/dev/null + + # This doesn't work correctly. It creates: + # Number Start End Size File system Name Flags + # 1 1049kB 24.1MB 23.1MB boot msftdata + # Odd if the same is entered interactively it does set the FS type + parted -s -a optimal -- "${IMG}" mkpart boot fat32 1MiB 23MiB + + # Map this partition to a loop device + kp="$(sudo kpartx -av ${IMG})" + read boot_dev<<<$(grep -o 'loop.*p.' <<< "${kp}") + test "${boot_dev}" + dev="/dev/mapper/${boot_dev}" + + mkfs.vfat "${dev}" >/dev/null + + sudo mount -o loop "${dev}" $MNT + + copy_files + + # Sync here since this makes kpartx more likely to work the first time + sync + sudo umount $MNT + + # For some reason this needs a sleep or it sometimes fails, if it was + # run recently (in the last few seconds) + if ! sudo kpartx -d "${IMG}" > /dev/null; then + sleep .5 + sudo kpartx -d "${IMG}" > /dev/null || \ + echo "Failed to remove ${boot_dev}, use: sudo kpartx -d ${IMG}" + fi +} + +TMP="/tmp/efi${bitness}${type}" +MNT=/mnt/x +BUILD="efi-x86_${type}${bitness}" +IMG=try.img + +if [[ -n "${old}" && "${bitness}" = "32" ]]; then + BUILD="efi-x86_${type}" +fi + +setup_files + +qemu-img create "${IMG}" 24M >/dev/null + +if [[ -n "${part}" ]]; then + setup_part +else + setup_raw +fi + +if [[ -n "${run}" ]]; then + run_qemu +fi
Re: [PATCH v2 07/39] efi: Add a script for building and testing U-Boot on UEFI
Hi Heinrich,
Quoted 19 lines · Heinrich Schuchardt
Quoted 7 lines · Simon Glass
It is quite complicating to run U-Boot on qemu since we have four different builds and they must use different versions of qemu and the UEFI binaries.
Add a script to help.
Signed-off-by Simon Glass <[email protected]>This patch seems not to work:
$ scripts/build-efi.sh -wsPr Packaging efi-x86_app32 cp: cannot stat '/tmp/b/efi-x86_app32/u-boot-app.efi': No such file or director
Quoted 4 lines
The scripts lacks a step to build the EFI binary.
If you expect the user to build it, you will find it in the current directory and not in some arbitrary subdirectory of /tmp.
Did you read the top of the script? It does not do the build, just the package/run bit
Quoted 3 lines
Temporary directories should be created with 'mktemp -d'. Using fixed paths (e.g.
/tmp/b/) may lead to conflicts.
I'll make it a variable in v3. But we cannot use mktemp-d as the build needs to stay around for multiple runs of this script.
Regards,
Simon
PATCH v2 8/39 efi: Enable DM_ETH for the app
In reply to Simon GlassThere is no need to avoid driver model for networking. Drop this.
---
Changes in v2:
- Add new patch to enable DM_ETH for the app
configs/efi-x86_app_defconfig | 1 - 1 file changed, 1 deletion(-)
configs/efi-x86_app_defconfig +0 -1
@@ -32,7 +32,6 @@ CONFIG_ENV_OVERWRITE=y CONFIG_SYS_RELOC_GD_ENV_ADDR=y CONFIG_REGMAP=y CONFIG_SYSCON=y -# CONFIG_DM_ETH is not set # CONFIG_REGEX is not set # CONFIG_GZIP is not set CONFIG_EFI=y
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 9/39 efi: Drop the OF_EMBED warning for EFI
In reply to Simon GlassFor the EFI app, we must embed the devicetree in the ELF file since that is the only thing that is run by UEFI. Drop the warning to avoid confusion.
---
Changes in v2:
- Add new patch to drop the OF_EMBED warning for EFI
Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Makefile +1 -1
@@ -1087,7 +1087,7 @@ endif ifeq ($(CONFIG_DEPRECATED),y) $(warning "You have deprecated configuration options enabled in your .config! Please check your configuration.") endif -ifeq ($(CONFIG_OF_EMBED),y) +ifeq ($(CONFIG_OF_EMBED)$(CONFIG_EFI_APP),y) @echo >&2 "===================== WARNING ======================" @echo >&2 "CONFIG_OF_EMBED is enabled. This option should only" @echo >&2 "be used for debugging purposes. Please use"
Signature · 1 line
2.33.0.685.g46640cef36-goog
Re: [PATCH v2 09/39] efi: Drop the OF_EMBED warning for EFI
Quoted 18 lines · Simon Glass
For the EFI app, we must embed the devicetree in the ELF file since that is the only thing that is run by UEFI. Drop the warning to avoid confusion.
Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- Add new patch to drop theOF_EMBEDwarning for EFIMakefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)Makefile +0
-0
Makefile endif · line 1093
Quoted 8 lines
Makefile +1
-1@@ -1087,7 +1087,7 @@ endif ifeq ($(CONFIG_DEPRECATED),y) $(warning "You have deprecated configuration options enabled in your .config! Please check your configuration.") endif -ifeq ($(CONFIG_OF_EMBED),y) +ifeq ($(CONFIG_OF_EMBED)$(CONFIG_EFI_APP),y) @echo >&2 "===================== WARNING ======================" @echo >&2 "CONFIG_OF_EMBED is enabled. This option should only" @echo >&2 "be used for debugging purposes. Please use"
There are 47 defconfigs and 5 include/configs/*.h using CONFIG_OF_EMBED. doc/board/intel/slimbootloader.rst and doc/board/coreboot/coreboot.rst require to use it for building U-Boot.
Is there a migration requirement and plan?
Best regards
Heinrich
Re: [PATCH v2 09/39] efi: Drop the OF_EMBED warning for EFI
Hi Heinrich,
Quoted 35 lines · Heinrich Schuchardt
Quoted 18 lines · Simon Glass
For the EFI app, we must embed the devicetree in the ELF file since that is the only thing that is run by UEFI. Drop the warning to avoid confusion.
Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- Add new patch to drop theOF_EMBEDwarning for EFIMakefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)Makefile +0
-0Makefile endif · line 1093
Quoted 8 lines
Makefile +1
-1@@ -1087,7 +1087,7 @@ endif ifeq ($(CONFIG_DEPRECATED),y) $(warning "You have deprecated configuration options enabled in your .config! Please check your configuration.") endif -ifeq ($(CONFIG_OF_EMBED),y) +ifeq ($(CONFIG_OF_EMBED)$(CONFIG_EFI_APP),y) @echo >&2 "===================== WARNING ======================" @echo >&2 "CONFIG_OF_EMBED is enabled. This option should only" @echo >&2 "be used for debugging purposes. Please use"There are 47 defconfigs and 5
include/configs/*.husingCONFIG_OF_EMBED.doc/board/intel/slimbootloader.rstanddoc/board/coreboot/coreboot.rstrequire to use it for building U-Boot.
Quoted 1 line
Is there a migration requirement and plan?
Actually people were never supposed to use it.
I sent a series today which cleans this up quite a bit, so that I think many of the boards using OF_EMBED can drop it. TBD though.
Regards,
Simon
PATCH v2 10/39 x86: Create a 32/64-bit selection for the app
In reply to Simon GlassMost EFI implementations use 64-bit but U-Boot only supports running as a 32-bit app at present. While efi-x86_payload64 does boot from 64-bit UEFI it immediately changes back to 32-bit before starting U-Boot.
In order to support a 64-bit U-Boot app, update the Kconfig to add an option for 32/64 bit. Update the prompt for the existing option so it is clear it relates to the stub. Move both up to just under the choice that controls them, since this looks better and the menu.
Use CONFIG_EFI_APP in the Makefile instead of CONFIG_TARGET_EFI_APP, since the latter is specific to a single target and we will have two.
Memory size is set to 32MB for now so that it can run on qemu without increasing the default memory size. We may need to increase the default later.
---
Changes in v2:
- Expand the commit message to make things clearer
arch/x86/cpu/intel_common/Makefile | 2 +- lib/efi/Kconfig | 34 +++++++++++++++++++++++------- 2 files changed, 27 insertions(+), 9 deletions(-)
arch/x86/cpu/intel_common/Makefile +1 -1
@@ -27,7 +27,7 @@ obj-y += fast_spi.o obj-y += lpc.o obj-y += lpss.o obj-$(CONFIG_$(SPL_)INTEL_GENERIC_WIFI) += generic_wifi.o -ifndef CONFIG_TARGET_EFI_APP +ifndef CONFIG_EFI_APP obj-$(CONFIG_$(SPL_TPL_)X86_32BIT_INIT) += microcode.o ifndef CONFIG_$(SPL_)X86_64 obj-y += microcode.o
lib/efi/Kconfig +26 -8
@@ -26,18 +26,26 @@ config EFI_STUB endchoice -config EFI_RAM_SIZE - hex "Amount of EFI RAM for U-Boot" +choice + prompt "EFI app 32/64-bit selection" depends on EFI_APP - default 0x2000000 help - Set the amount of EFI RAM which is claimed by U-Boot for its own - use. U-Boot allocates this from EFI on start-up (along with a few - other smaller amounts) and it can never be increased after that. - It is used as the RAM size in with U-Boot. + EFI does not support mixing 32-bit and 64-bit modes. This is a + significant problem because it means that you must build a stub with + the correct type for EFI to load it correctly. If you are using + 32-bit EFI, select 32-bit here, else select 64-bit. Failure to do + this may produce no error message - it just won't start! + +config EFI_APP_32BIT + bool "Produce an app for running with 32-bit EFI" + +config EFI_APP_64BIT + bool "Produce an app for running with 64-bit EFI" + +endchoice choice - prompt "EFI 32/64-bit selection" + prompt "EFI stub 32/64-bit selection" depends on EFI_STUB help EFI does not support mixing 32-bit and 64-bit modes. This is a
@@ -53,3 +61,13 @@ config EFI_STUB_64BIT bool "Produce a stub for running with 64-bit EFI" endchoice + +config EFI_RAM_SIZE + hex "Amount of EFI RAM for U-Boot" + depends on EFI_APP + default 0x2000000 + help + Set the amount of EFI RAM which is claimed by U-Boot for its own + use. U-Boot allocates this from EFI on start-up (along with a few + other smaller amounts) and it can never be increased after that. + It is used as the RAM size in with U-Boot.
Signature · 1 line
2.33.0.685.g46640cef36-goog
Re: [PATCH v2 10/39] x86: Create a 32/64-bit selection for the app
Quoted 79 lines · Simon Glass
Most EFI implementations use 64-bit but U-Boot only supports running as a 32-bit app at present. While efi-x86_payload64 does boot from 64-bit UEFI it immediately changes back to 32-bit before starting U-Boot.
In order to support a 64-bit U-Boot app, update the
Kconfigto add an option for 32/64 bit. Update the prompt for the existing option so it is clear it relates to the stub. Move both up to just under the choice that controls them, since this looks better and the menu.Use
CONFIG_EFI_APPin theMakefileinstead ofCONFIG_TARGET_EFI_APP, since the latter is specific to a single target and we will have two.Memory size is set to 32MB for now so that it can run on qemu without increasing the default memory size. We may need to increase the default later.
Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- Expand the commit message to make things clearerarch/x86/cpu/intel_common/Makefile | 2 +- lib/efi/Kconfig | 34 +++++++++++++++++++++++------- 2 files changed, 27 insertions(+), 9 deletions(-)arch/x86/cpu/intel_common/Makefile +1
-1@@ -27,7 +27,7 @@ obj-y += fast_spi.o obj-y += lpc.o obj-y += lpss.o obj-$(CONFIG_$(SPL_)INTEL_GENERIC_WIFI) += generic_wifi.o -ifndef CONFIG_TARGET_EFI_APP +ifndef CONFIG_EFI_APP obj-$(CONFIG_$(SPL_TPL_)X86_32BIT_INIT) += microcode.o ifndef CONFIG_$(SPL_)X86_64 obj-y += microcode.olib/efi/Kconfig +16
-8@@ -26,18 +26,26 @@ config EFI_STUB endchoice -config EFI_RAM_SIZE - hex "Amount of EFI RAM for U-Boot" +choice + prompt "EFI app 32/64-bit selection" depends on EFI_APP - default 0x2000000 help - Set the amount of EFI RAM which is claimed by U-Boot for its own - use. U-Boot allocates this from EFI on start-up (along with a few - other smaller amounts) and it can never be increased after that. - It is used as the RAM size in with U-Boot. + EFI does not support mixing 32-bit and 64-bit modes. This is a + significant problem because it means that you must build a stub with + the correct type for EFI to load it correctly. If you are using + 32-bit EFI, select 32-bit here, else select 64-bit. Failure to do + this may produce no error message - it just won't start! + +config EFI_APP_32BIT + bool "Produce an app for running with 32-bit EFI" + +config EFI_APP_64BIT + bool "Produce an app for running with 64-bit EFI" + +endchoice choice - prompt "EFI 32/64-bit selection" + prompt "EFI stub 32/64-bit selection" depends on EFI_STUB help EFI does not support mixing 32-bit and 64-bit modes. This is a
lib/efi/Kconfig config EFI_STUB_64BIT · line 68
Quoted 8 lines
lib/efi/Kconfig +5
-0@@ -53,3 +61,13 @@ config EFI_STUB_64BIT bool "Produce a stub for running with 64-bit EFI" endchoice + +config EFI_RAM_SIZE + hex "Amount of EFI RAM for U-Boot" + depends on EFI_APP + default 0x2000000
This will be too small for loading the kernel and the initrd in some cases. 256 MiB seems reasonable.
Best regards
Heinrich
Quoted 6 lines
lib/efi/Kconfig +5
-0+ help + Set the amount of EFI RAM which is claimed by U-Boot for its own + use. U-Boot allocates this from EFI on start-up (along with a few + other smaller amounts) and it can never be increased after that. + It is used as the RAM size in with U-Boot.
PATCH v2 11/39 efi: Create a 64-bit app
In reply to Simon GlassMost modern platforms use 64-bit EFI so it is useful to have a U-Boot app that runs under that. Add a (non-functional) build for this.
Note that --whole-archive causes the gcc 9.2 linker to crash, so disable this for now. Once this is resolved, things should work.
For now, avoid mentioning the documentation for the 64-bit app, since it does not work.
---
Changes in v2:
- Add MAINTAINERS entry
- Add a work-around to avoid a toolchain crash
MAINTAINERS | 3 ++
Makefile | 8 +++-
arch/x86/cpu/u-boot-64.lds | 2 +
arch/x86/cpu/x86_64/Makefile | 4 ++
arch/x86/cpu/x86_64/cpu.c | 17 --------
arch/x86/cpu/x86_64/misc.c | 25 ++++++++++++
board/efi/Kconfig | 15 ++++++-
board/efi/efi-x86_app/Kconfig | 2 +-
board/efi/efi-x86_app/MAINTAINERS | 11 +++++-
..._app_defconfig => efi-x86_app32_defconfig} | 2 +-
configs/efi-x86_app64_defconfig | 39 +++++++++++++++++++
doc/develop/uefi/u-boot_on_efi.rst | 6 +--
12 files changed, 106 insertions(+), 28 deletions(-)
create mode 100644 arch/x86/cpu/x86_64/misc.c
rename configs/{efi-x86_app_defconfig => efi-x86_app32_defconfig} (97%)
create mode 100644 configs/efi-x86_app64_defconfigMAINTAINERS +3 -0
@@ -705,6 +705,9 @@ M: Simon Glass <[email protected]> M: Heinrich Schuchardt <[email protected]> S: Maintained W: https://u-boot.readthedocs.io/en/latest/develop/uefi/u-boot_on_efi.html +F: board/efi/efi-x86_app +F: configs/efi-x86_app* +F: doc/develop/uefi/u-boot_on_efi.rst F: lib/efi/efi_app.c F: scripts/build-efi.sh
Makefile +6 -2
@@ -1755,12 +1755,16 @@ quiet_cmd_u-boot__ ?= LTO $@ -Wl,-Map,u-boot.map; \ $(if $(ARCH_POSTLINK), $(MAKE) -f $(ARCH_POSTLINK) $@, true) else +# Note: Linking efi-x86_app64 causes a segfault in the linker at present +# when using x86_64-linux-gnu-ld.bfd +# For now, disable --whole-archive which makes things link, although not +# correctly quiet_cmd_u-boot__ ?= LD $@ cmd_u-boot__ ?= $(LD) $(KBUILD_LDFLAGS) $(LDFLAGS_u-boot) -o $@ \ -T u-boot.lds $(u-boot-init) \ - --whole-archive \ + $(if $(CONFIG_EFI_APP_64BIT),,--whole-archive) \ $(u-boot-main) \ - --no-whole-archive \ + $(if $(CONFIG_EFI_APP_64BIT),,--no-whole-archive) \ $(PLATFORM_LIBS) -Map u-boot.map; \ $(if $(ARCH_POSTLINK), $(MAKE) -f $(ARCH_POSTLINK) $@, true) endif
arch/x86/cpu/u-boot-64.lds +2 -0
@@ -15,7 +15,9 @@ SECTIONS /DISCARD/ : { *(.u_boot_list_2_cmd_*) } #endif +#ifdef CONFIG_SYS_TEXT_BASE . = CONFIG_SYS_TEXT_BASE; /* Location of bootcode in flash */ +#endif __text_start = .; .text.start : { *(.text.start); }
arch/x86/cpu/x86_64/Makefile +4 -0
@@ -4,3 +4,7 @@ # obj-y += cpu.o interrupts.o setjmp.o + +ifndef CONFIG_EFI +obj-y += misc.o +endif
arch/x86/cpu/x86_64/cpu.c +0 -17
@@ -52,23 +52,6 @@ int x86_mp_init(void)return 0 ; } -int misc_init_r (void ) -{ -return 0 ; -} - -#ifndef CONFIG_SYS_COREBOOT -int checkcpu (void ) -{ -return 0 ; -} - -int print_cpuinfo (void ) -{ -return 0 ; -} -#endif -int x86_cpu_reinit_f (void ) {return 0 ;
arch/x86/cpu/x86_64/misc.c new file +25 -0
@@ -0,0 +1,25 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* (C) Copyright 2016 Google, Inc +* Written by Simon Glass <[email protected]> +*/ + +#include <common.h> +#include <init.h> + +int misc_init_r (void ) +{ +return 0 ; +} + +#ifndef CONFIG_SYS_COREBOOT +int checkcpu (void ) +{ +return 0 ; +} + +int print_cpuinfo (void ) +{ +return 0 ; +} +#endif
board/efi/Kconfig +13 -2
@@ -4,14 +4,25 @@ choice prompt "Mainboard model" optional -config TARGET_EFI_APP - bool "efi application" +config TARGET_EFI_APP32 + bool "32-bit efi application" + select EFI_APP help This target is used for running U-Boot on top of EFI. In this case EFI does the early initialisation, and U-Boot takes over once the RAM, video and CPU are fully running. U-Boot is loaded as an application from EFI. +config TARGET_EFI_APP64 + bool "64-bit efi application" + select EFI_APP + select X86_64 + help + This target is used for running U-Boot on top of EFI in 64-bit mode. + In this case EFI does the early initialisation, and U-Boot + takes over once the RAM, video and CPU are fully running. + U-Boot is loaded as an application from EFI. + config TARGET_EFI_PAYLOAD bool "efi payload" help
board/efi/efi-x86_app/Kconfig +1 -1
@@ -1,4 +1,4 @@ -if TARGET_EFI_APP +if EFI_APP config SYS_BOARD default "efi-x86_app"
board/efi/efi-x86_app/MAINTAINERS +9 -2
@@ -1,6 +1,13 @@ -EFI-X86_APP BOARD +EFI-X86_APP32 BOARD M: Simon Glass <[email protected]> S: Maintained F: board/efi/efi-x86_app/ F: include/configs/efi-x86_app.h -F: configs/efi-x86_app_defconfig +F: configs/efi-x86_app32_defconfig + +EFI-X86_APP64 BOARD +M: Simon Glass <[email protected]> +S: Maintained +F: board/efi/efi-x86_app/ +F: include/configs/efi-x86_app.h +F: configs/efi-x86_app64_defconfig
configs/efi-x86_app32_defconfig renamed from configs/efi-x86_app_defconfig +1 -1
@@ -5,7 +5,7 @@ CONFIG_DEFAULT_DEVICE_TREE="efi-x86_app" CONFIG_DEBUG_UART_BASE=0 CONFIG_DEBUG_UART_CLOCK=0 CONFIG_VENDOR_EFI=y -CONFIG_TARGET_EFI_APP=y +CONFIG_TARGET_EFI_APP32=y CONFIG_DEBUG_UART=y CONFIG_FIT=y CONFIG_SHOW_BOOT_PROGRESS=y
configs/efi-x86_app64_defconfig new file +39 -0
@@ -0,0 +1,39 @@ +CONFIG_X86=y +CONFIG_NR_DRAM_BANKS=8 +CONFIG_ENV_SIZE=0x1000 +CONFIG_DEFAULT_DEVICE_TREE="efi-x86_app" +CONFIG_DEBUG_UART_BASE=0 +CONFIG_DEBUG_UART_CLOCK=0 +CONFIG_VENDOR_EFI=y +CONFIG_TARGET_EFI_APP64=y +CONFIG_DEBUG_UART=y +CONFIG_FIT=y +CONFIG_SHOW_BOOT_PROGRESS=y +CONFIG_USE_BOOTARGS=y +CONFIG_BOOTARGS="root=/dev/sdb3 init=/sbin/init rootwait ro" +CONFIG_SYS_CONSOLE_INFO_QUIET=y +CONFIG_DISPLAY_BOARDINFO_LATE=y +CONFIG_LAST_STAGE_INIT=y +CONFIG_HUSH_PARSER=y +# CONFIG_CMD_BOOTM is not set +CONFIG_CMD_PART=y +# CONFIG_CMD_NET is not set +CONFIG_CMD_TIME=y +CONFIG_CMD_EXT2=y +CONFIG_CMD_EXT4=y +CONFIG_CMD_EXT4_WRITE=y +CONFIG_CMD_FAT=y +CONFIG_CMD_FS_GENERIC=y +CONFIG_MAC_PARTITION=y +CONFIG_ISO_PARTITION=y +CONFIG_EFI_PARTITION=y +CONFIG_OF_EMBED=y +CONFIG_ENV_OVERWRITE=y +CONFIG_SYS_RELOC_GD_ENV_ADDR=y +CONFIG_REGMAP=y +CONFIG_SYSCON=y +# CONFIG_REGEX is not set +# CONFIG_GZIP is not set +CONFIG_EFI=y +CONFIG_EFI_APP_64BIT=y +# CONFIG_EFI_LOADER is not set
doc/develop/uefi/u-boot_on_efi.rst +3 -3
@@ -48,10 +48,10 @@ for that board. It will be either 32-bit or 64-bit. Alternatively, you can opt for using QEMU [1] and the OVMF [2], as detailed below. To build U-Boot as an EFI application (32-bit EFI required), enable CONFIG_EFI -and CONFIG_EFI_APP. The efi-x86_app config (efi-x86_app_defconfig) is set up +and CONFIG_EFI_APP. The efi-x86_app config (efi-x86_app32_defconfig) is set up for this. Just build U-Boot as normal, e.g.:: - make efi-x86_app_defconfig + make efi-x86_app32_defconfig make To build U-Boot as an EFI payload (32-bit or 64-bit EFI can be used), enable
@@ -259,7 +259,7 @@ This work could be extended in a number of ways: - Add ARM support -- Add 64-bit application support +- Add 64-bit application support (in progress) - Figure out how to solve the interrupt problem
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 12/39 x86: Don't duplicate global_ptr in 64-bit EFI app
In reply to Simon GlassThis variable is already defined by the EFI code. Drop the duplicate definition when building a 64-bit EFI app.
---
(no changes since v1)
arch/x86/cpu/x86_64/cpu.c | 16 ---------------- arch/x86/cpu/x86_64/misc.c | 16 ++++++++++++++++ lib/efi/efi.c | 9 +++++++++ 3 files changed, 25 insertions(+), 16 deletions(-)
arch/x86/cpu/x86_64/cpu.c +0 -16
@@ -10,22 +10,6 @@#include <init.h> #include <asm/global_data.h> -DECLARE_GLOBAL_DATA_PTR ; - -/* -* Global declaration of gd. -* -* As we write to it before relocation we have to make sure it is not put into -* a .bss section which may overlap a .rela section. Initialization forces it -* into a .data section which cannot overlap any .rela section. -*/ -struct global_data *global_data_ptr = (struct global_data *)~0 ; - -void arch_setup_gd (gd_t *new_gd) -{ - global_data_ptr = new_gd; -} -int cpu_has_64bit (void ) {return true;
arch/x86/cpu/x86_64/misc.c +16 -0
@@ -7,6 +7,22 @@#include <common.h> #include <init.h> +DECLARE_GLOBAL_DATA_PTR ; + +/* +* Global declaration of gd. +* +* As we write to it before relocation we have to make sure it is not put into +* a .bss section which may overlap a .rela section. Initialization forces it +* into a .data section which cannot overlap any .rela section. +*/ +struct global_data *global_data_ptr = (struct global_data *)~0 ; + +void arch_setup_gd (gd_t *new_gd) +{ + global_data_ptr = new_gd; +} +int misc_init_r (void ) {return 0 ;
lib/efi/efi.c +9 -0
@@ -17,6 +17,15 @@#include <efi.h> #include <efi_api.h> +/* +* Global declaration of gd. +* +* As we write to it before relocation we have to make sure it is not put into +* a .bss section which may overlap a .rela section. Initialization forces it +* into a .data section which cannot overlap any .rela section. +*/ +struct global_data *global_data_ptr = (struct global_data *)~0 ; +/* * Unfortunately we cannot access any code outside what is built especially * for the stub. lib/string.c is already being built for the U-Boot payload
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 13/39 efi: Add a way to obtain boot services in the app
In reply to Simon GlassAdd a function to return this information along with a stub for the efi_info_get() function, since calling it otherwise hangs U-Boot.
---
(no changes since v1)
include/efi.h | 8 +++++++- lib/efi/efi_app.c | 10 ++++++++++ 2 files changed, 17 insertions(+), 1 deletion(-)
include/efi.h +7 -1
@@ -444,9 +444,15 @@ extern char _binary_u_boot_bin_start[], _binary_u_boot_bin_end[];* * @return pointer to EFI system table */ -struct efi_system_table *efi_get_sys_table (void ); +/** +* efi_get_boot() - Get access to the EFI boot services table +* +* @return pointer to EFI boot services table +*/ +struct efi_boot_services *efi_get_boot (void ); +/** * efi_get_ram_base() - Find the base of RAM *
lib/efi/efi_app.c +10 -0
@@ -31,11 +31,21 @@ struct efi_system_table *efi_get_sys_table(void)return global_priv->sys_table; } +struct efi_boot_services *efi_get_boot (void ) +{ +return global_priv->boot; +} +unsigned long efi_get_ram_base (void ) {return global_priv->ram_base; } +int efi_info_get (enum efi_entry_t type,void **datap,int *sizep) +{ +return -ENOSYS ; +} +static efi_status_t setup_memory (struct efi_priv *priv) {struct efi_boot_services *boot = priv->boot;
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 14/39 efi: Add video support to the app
In reply to Simon GlassThe current EFI video driver only works when running in the stub. In that case the stub calls boot services (before jumping to U-Boot proper) and copies the graphics info over to the efi table. This is necessary because the stub exits boot services before jumping to U-Boot.
The app maintains access to boot services throughout its life, so does not need to do this. Update the driver to support calling boot services directly.
Enable video output for the app. Note that this uses the EFI_GRAPHICS_OUTPUT_PROTOCOL protocol, even though it mentions vesa.
A sample qemu command-line for this case is:
qemu-system-x86_64 -bios /usr/share/edk2.git/ovmf-ia32/OVMF-pure-efi.fd -drive id=disk,file=try.img,if=none,format=raw -nic none -device ahci,id=ahci -device ide-hd,drive=disk,bus=ahci.0
---
Changes in v2:
- Add a note that EFI_GRAPHICS_OUTPUT_PROTOCOL is used
- Update documentation
arch/x86/dts/efi-x86_app.dts | 4 +++ board/efi/efi-x86_app/Kconfig | 4 +++ doc/develop/uefi/u-boot_on_efi.rst | 2 +- drivers/video/Kconfig | 2 +- drivers/video/efi.c | 45 ++++++++++++++++++++++++------ include/configs/efi-x86_app.h | 6 ++-- 6 files changed, 50 insertions(+), 13 deletions(-)
arch/x86/dts/efi-x86_app.dts +4 -0
@@ -25,4 +25,8 @@ compatible = "efi,reset"; u-boot,dm-pre-reloc; }; + efi-fb { + compatible = "efi-fb"; + }; + };
board/efi/efi-x86_app/Kconfig +4 -0
@@ -12,4 +12,8 @@ config SYS_SOC config SYS_CONFIG_NAME default "efi-x86_app" +config BOARD_SPECIFIC_OPTIONS # dummy + def_bool y + imply VIDEO_EFI + endif
doc/develop/uefi/u-boot_on_efi.rst +1 -1
@@ -263,7 +263,7 @@ This work could be extended in a number of ways: - Figure out how to solve the interrupt problem -- Add more drivers to the application side (e.g. video, block devices, USB, +- Add more drivers to the application side (e.g. block devices, USB, environment access). This would mostly be an academic exercise as a strong use case is not readily apparent, but it might be fun.
drivers/video/Kconfig +1 -1
@@ -250,7 +250,7 @@ config VIDEO_COREBOOT config VIDEO_EFI bool "Enable EFI framebuffer driver support" - depends on EFI_STUB + depends on EFI_STUB || EFI_APP help Turn on this option to enable a framebuffeer driver when U-Boot is loaded as a payload (see README.u-boot_on_efi) by an EFI BIOS where
drivers/video/efi.c +37 -8
59 lines not on this page · raw message
include/configs/efi-x86_app.h +3 -3
@@ -10,8 +10,8 @@#undef CONFIG_TPM_TIS_BASE_ADDRESS -#define CONFIG_STD_DEVICES_SETTINGS "stdin=usbkbd,vga,serial\0" \ - "stdout=vga,serial\0" \ - "stderr=vga,serial\0" +#define CONFIG_STD_DEVICES_SETTINGS "stdin=serial\0" \ + "stdout=vidconsole\0" \ + "stderr=vidconsole\0"#endif
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 15/39 efi: Add EFI uclass for media
In reply to Simon GlassAt present UCLASS_EFI is used to represent an EFI filesystem among other things. The description of this uclass is "EFI managed devices" which is pretty vague. The only driver that uses this uclass is in fact not a real U-Boot driver, since its operations do not include a struct udevice.
Rather than mess with this, create a new UCLASS_EFI_MEDIA uclass to handle EFI media such as disks. Make this the uclass to use for EFI media so that it can be used with 'part list', for example.
The existing implementation using UCLASS_EFI remains as is, for discussion.
---
Changes in v2:
- Add MAINTAINERS entry
- Show the correct interface type with 'part list'
- Update the commit message to explain things better
MAINTAINERS | 3 +++ arch/sandbox/dts/test.dts | 4 ++++ disk/part.c | 5 ++++- drivers/block/Kconfig | 23 +++++++++++++++++++++++ drivers/block/Makefile | 3 +++ drivers/block/blk-uclass.c | 2 +- drivers/block/efi-media-uclass.c | 15 +++++++++++++++ drivers/block/sb_efi_media.c | 20 ++++++++++++++++++++ include/dm/uclass-id.h | 1 + test/dm/Makefile | 1 + test/dm/efi_media.c | 24 ++++++++++++++++++++++++ 11 files changed, 99 insertions(+), 2 deletions(-) create mode 100644 drivers/block/efi-media-uclass.c create mode 100644 drivers/block/sb_efi_media.c create mode 100644 test/dm/efi_media.c
MAINTAINERS +3 -0
@@ -708,8 +708,11 @@ W: https://u-boot.readthedocs.io/en/latest/develop/uefi/u-boot_on_efi.html F: board/efi/efi-x86_app F: configs/efi-x86_app* F: doc/develop/uefi/u-boot_on_efi.rst +F: drivers/block/efi-media-uclass.c +F: drivers/block/sb_efi_media.c F: lib/efi/efi_app.c F: scripts/build-efi.sh +F: test/dm/efi_media.c EFI PAYLOAD M: Heinrich Schuchardt <[email protected]>
arch/sandbox/dts/test.dts +4 -0
@@ -495,6 +495,10 @@ compatible = "sandbox,clk-ccf"; }; + efi-media { + compatible = "sandbox,efi-media"; + }; + eth@10002000 { compatible = "sandbox,eth"; reg = <0x10002000 0x1000>;
disk/part.c +4 -1
@@ -296,8 +296,11 @@ static void print_part_header(const char *type, struct blk_desc *dev_desc)case IF_TYPE_VIRTIO :puts ("VirtIO" );break ; +case IF_TYPE_EFI : +puts ("EFI" ); +break ;default : -puts ("UNKNOWN" ); +puts ("UNKNOWN" );break ; } printf (" device %d -- Partition Type: %s\n\n",
drivers/block/Kconfig +23 -0
@@ -61,6 +61,29 @@ config TPL_BLOCK_CACHE help This option enables the disk-block cache in TPL +config EFI_MEDIA + bool "Support EFI media drivers" + default y if EFI || SANDBOX + help + Enable this to support media devices on top of UEFI. This enables + just the uclass so you also need a specific driver to make this do + anything. + + For sandbox there is a test driver. + +if EFI_MEDIA + +config EFI_MEDIA_SANDBOX + bool "Sandbox EFI media driver" + depends on SANDBOX + default y + help + Enables a simple sandbox media driver, used for testing just the + EFI_MEDIA uclass. It does not do anything useful, since sandbox does + not actually support running on top of UEFI. + +endif # EFI_MEDIA + config IDE bool "Support IDE controllers" select HAVE_BLOCK_DEVICE
drivers/block/Makefile +3 -0
@@ -14,3 +14,6 @@ obj-$(CONFIG_IDE) += ide.o endif obj-$(CONFIG_SANDBOX) += sandbox.o obj-$(CONFIG_$(SPL_TPL_)BLOCK_CACHE) += blkcache.o + +obj-$(CONFIG_EFI_MEDIA) += efi-media-uclass.o +obj-$(CONFIG_EFI_MEDIA_SANDBOX) += sb_efi_media.o
drivers/block/blk-uclass.c +1 -1
@@ -44,7 +44,7 @@ static enum uclass_id if_type_uclass_id[IF_TYPE_COUNT] = { [IF_TYPE_SATA] = UCLASS_AHCI, [IF_TYPE_HOST] = UCLASS_ROOT, [IF_TYPE_NVME] = UCLASS_NVME, - [IF_TYPE_EFI] = UCLASS_EFI, + [IF_TYPE_EFI] = UCLASS_EFI_MEDIA, [IF_TYPE_VIRTIO] = UCLASS_VIRTIO, [IF_TYPE_PVBLOCK] = UCLASS_PVBLOCK, };
drivers/block/efi-media-uclass.c new file +15 -0
@@ -0,0 +1,15 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* Uclass for EFI media devices +* +* Copyright 2021 Google LLC +*/ + +#include <common.h> +#include <dm.h> + +UCLASS_DRIVER (efi_media) = { + .id =UCLASS_EFI_MEDIA , + .name ="efi_media" , + .flags =DM_UC_FLAG_SEQ_ALIAS , +};
drivers/block/sb_efi_media.c new file +20 -0
@@ -0,0 +1,20 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* EFI_MEDIA driver for sandbox +* +* Copyright 2021 Google LLC +*/ + +#include <common.h> +#include <dm.h> + +static const struct udevice_id sandbox_efi_media_ids[] = { + { .compatible ="sandbox,efi-media" }, + { } +}; + +U_BOOT_DRIVER (sandbox_efi_media) = { + .name ="sandbox_efi_media" , + .id =UCLASS_EFI_MEDIA , + .of_match = sandbox_efi_media_ids, +};
include/dm/uclass-id.h +1 -0
@@ -49,6 +49,7 @@ enum uclass_id {UCLASS_DSA ,/* Distributed (Ethernet) Switch Architecture */ UCLASS_ECDSA ,/* Elliptic curve cryptographic device */ UCLASS_EFI ,/* EFI managed devices */ +UCLASS_EFI_MEDIA ,/* EFI media (e.g. a disk) */ UCLASS_ETH ,/* Ethernet device */ UCLASS_ETH_PHY ,/* Ethernet PHY device */ UCLASS_FIRMWARE, /* Firmware */
test/dm/Makefile +1 -0
@@ -37,6 +37,7 @@ obj-$(CONFIG_DMA) += dma.o obj-$(CONFIG_VIDEO_MIPI_DSI) += dsi_host.o obj-$(CONFIG_DM_DSA) += dsa.o obj-$(CONFIG_ECDSA_VERIFY) += ecdsa.o +obj-$(CONFIG_EFI_MEDIA_SANDBOX) += efi_media.o obj-$(CONFIG_DM_ETH) += eth.o ifneq ($(CONFIG_EFI_PARTITION),) obj-$(CONFIG_FASTBOOT_FLASH_MMC) += fastboot.o
test/dm/efi_media.c new file +24 -0
@@ -0,0 +1,24 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* Test for EFI_MEDIA uclass +* +* Copyright 2021 Google LLC +*/ + +#include <common.h> +#include <dm.h> +#include <asm/test.h> +#include <dm/test.h> +#include <test/test.h> +#include <test/ut.h> + +/* Test that we can use the EFI_MEDIA uclass */ +static int dm_test_efi_media (struct unit_test_state *uts) +{ +struct udevice *dev; + +ut_assertok (uclass_first_device_err (UCLASS_EFI_MEDIA , &dev)); + +return 0 ; +} +DM_TEST (dm_test_efi_media,UT_TESTF_SCAN_PDATA |UT_TESTF_SCAN_FDT );
Signature · 1 line
2.33.0.685.g46640cef36-goog
Re: [PATCH v2 15/39] efi: Add EFI uclass for media
Quoted 1 line · Simon Glass
At present
UCLASS_EFIis used to represent an EFI filesystem among other
UCLASS_EFI is for UEFI drivers that provide the driver binding protocol.
Your new UCLASS seems to be for U-Boot drivers consuming the EFI_BLOCK_IO_PROTOCOL.
I agree that lib/efi_driver/efi_block_device.c is to much intertwined with the UEFI subsystem to easily merge it with your requirement. But it at least can serve as a template.
Quoted 6 lines
things. The description of this uclass is "EFI managed devices" which is pretty vague. The only driver that uses this uclass is in fact not a real U-Boot driver, since its operations do not include a struct udevice.
Rather than mess with this, create a new
UCLASS_EFI_MEDIAuclass to handle EFI media such as disks. Make this the uclass to use for EFI media so that
In UEFI speak MEDIA relates to:
* block devices
* firmware files
* Bluetooth
* REST services
* etc
Do you only mean block devices?
Why can't you use UCLASS_BLK for your new driver?
Best regards
Heinrich
Quoted 127 lines
it can be used with 'part list', for example.
The existing implementation using
UCLASS_EFIremains as is, for discussion.Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- AddMAINTAINERSentry
- Show the correct interface type with 'part list'
- Update the commit message to explain things betterMAINTAINERS | 3 +++ arch/sandbox/dts/test.dts | 4 ++++ disk/part.c | 5 ++++- drivers/block/Kconfig | 23 +++++++++++++++++++++++ drivers/block/Makefile | 3 +++ drivers/block/blk-uclass.c | 2 +- drivers/block/efi-media-uclass.c | 15 +++++++++++++++ drivers/block/sb_efi_media.c | 20 ++++++++++++++++++++ include/dm/uclass-id.h | 1 + test/dm/Makefile | 1 + test/dm/efi_media.c | 24 ++++++++++++++++++++++++ 11 files changed, 99 insertions(+), 2 deletions(-)create mode 100644
drivers/block/efi-media-uclass.ccreate mode 100644drivers/block/sb_efi_media.ccreate mode 100644test/dm/efi_media.cMAINTAINERS +3
-0@@ -708,8 +708,11 @@ W: https://u-boot.readthedocs.io/en/latest/develop/uefi/u-boot_on_efi.html F: board/efi/efi-x86_app F: configs/efi-x86_app* F: doc/develop/uefi/u-boot_on_efi.rst +F: drivers/block/efi-media-uclass.c +F: drivers/block/sb_efi_media.c F: lib/efi/efi_app.c F: scripts/build-efi.sh +F: test/dm/efi_media.c EFI PAYLOAD M: Heinrich Schuchardt <[email protected]>arch/sandbox/dts/test.dts +4
-0@@ -495,6 +495,10 @@ compatible = "sandbox,clk-ccf"; }; + efi-media { + compatible = "sandbox,efi-media"; + }; + eth@10002000 { compatible = "sandbox,eth"; reg = <0x10002000 0x1000>;disk/part.c +4
-1@@ -296,8 +296,11 @@ static void print_part_header(const char *type, struct blk_desc *dev_desc)case IF_TYPE_VIRTIO :puts ("VirtIO" );break ; +case IF_TYPE_EFI : +puts ("EFI" ); +break ;default : -puts ("UNKNOWN" ); +puts ("UNKNOWN" );break ; } printf (" device %d -- Partition Type: %s\n\n",drivers/block/Kconfig +23
-0@@ -61,6 +61,29 @@ config TPL_BLOCK_CACHE help This option enables the disk-block cache in TPL +config EFI_MEDIA + bool "Support EFI media drivers" + default y if EFI || SANDBOX + help + Enable this to support media devices on top of UEFI. This enables + just the uclass so you also need a specific driver to make this do + anything. + + For sandbox there is a test driver. + +if EFI_MEDIA + +config EFI_MEDIA_SANDBOX + bool "Sandbox EFI media driver" + depends on SANDBOX + default y + help + Enables a simple sandbox media driver, used for testing just the + EFI_MEDIA uclass. It does not do anything useful, since sandbox does + not actually support running on top of UEFI. + +endif # EFI_MEDIA + config IDE bool "Support IDE controllers" select HAVE_BLOCK_DEVICEdrivers/block/Makefile +3
-0@@ -14,3 +14,6 @@ obj-$(CONFIG_IDE) += ide.o endif obj-$(CONFIG_SANDBOX) += sandbox.o obj-$(CONFIG_$(SPL_TPL_)BLOCK_CACHE) += blkcache.o + +obj-$(CONFIG_EFI_MEDIA) += efi-media-uclass.o +obj-$(CONFIG_EFI_MEDIA_SANDBOX) += sb_efi_media.odrivers/block/blk-uclass.c +0
-0
drivers/block/blk-uclass.c static enum uclass_id if_type_uclass_id[IF_TYPE_COUNT] = { · line 47
Quoted 4 lines
drivers/block/blk-uclass.c +0
-1@@ -44,7 +44,7 @@ static enum uclass_id if_type_uclass_id[IF_TYPE_COUNT] = { [IF_TYPE_SATA] = UCLASS_AHCI, [IF_TYPE_HOST] = UCLASS_ROOT, [IF_TYPE_NVME] = UCLASS_NVME, - [IF_TYPE_EFI] = UCLASS_EFI,
UCLASS_EFI_MEDIA is only relevant for the EFI app and should not be compiled into normal U-Boot. This change is bound to break the UEFI sub-system.
Why don't you define a new if_type for your use case?
blk_get_devnum_by_typename() uses if_type to check if the corresponding uclass exists. Why is this check needed? If we remove it, we can drop if_type_uclass_id[] completely. This would be my preferred solution.
Best regards
Heinrich
Quoted 106 lines
drivers/block/blk-uclass.c +1
-0+ [IF_TYPE_EFI] = UCLASS_EFI_MEDIA, [IF_TYPE_VIRTIO] = UCLASS_VIRTIO, [IF_TYPE_PVBLOCK] = UCLASS_PVBLOCK, };drivers/block/efi-media-uclass.c new file +15
-0@@ -0,0 +1,15 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* Uclass for EFI media devices +* +* Copyright 2021 Google LLC +*/ + +#include <common.h> +#include <dm.h> + +UCLASS_DRIVER (efi_media) = { + .id =UCLASS_EFI_MEDIA , + .name ="efi_media" , + .flags =DM_UC_FLAG_SEQ_ALIAS , +};drivers/block/sb_efi_media.c new file +20
-0@@ -0,0 +1,20 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* EFI_MEDIA driver for sandbox +* +* Copyright 2021 Google LLC +*/ + +#include <common.h> +#include <dm.h> + +static const struct udevice_id sandbox_efi_media_ids[] = { + { .compatible ="sandbox,efi-media" }, + { } +}; + +U_BOOT_DRIVER (sandbox_efi_media) = { + .name ="sandbox_efi_media" , + .id =UCLASS_EFI_MEDIA , + .of_match = sandbox_efi_media_ids, +};include/dm/uclass-id.h +1
-0@@ -49,6 +49,7 @@ enum uclass_id {UCLASS_DSA ,/* Distributed (Ethernet) Switch Architecture */ UCLASS_ECDSA ,/* Elliptic curve cryptographic device */ UCLASS_EFI ,/* EFI managed devices */ +UCLASS_EFI_MEDIA ,/* EFI media (e.g. a disk) */ UCLASS_ETH ,/* Ethernet device */ UCLASS_ETH_PHY ,/* Ethernet PHY device */ UCLASS_FIRMWARE, /* Firmware */test/dm/Makefile +1
-0@@ -37,6 +37,7 @@ obj-$(CONFIG_DMA) += dma.o obj-$(CONFIG_VIDEO_MIPI_DSI) += dsi_host.o obj-$(CONFIG_DM_DSA) += dsa.o obj-$(CONFIG_ECDSA_VERIFY) += ecdsa.o +obj-$(CONFIG_EFI_MEDIA_SANDBOX) += efi_media.o obj-$(CONFIG_DM_ETH) += eth.o ifneq ($(CONFIG_EFI_PARTITION),) obj-$(CONFIG_FASTBOOT_FLASH_MMC) += fastboot.otest/dm/efi_media.c new file +24
-0@@ -0,0 +1,24 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* Test for EFI_MEDIA uclass +* +* Copyright 2021 Google LLC +*/ + +#include <common.h> +#include <dm.h> +#include <asm/test.h> +#include <dm/test.h> +#include <test/test.h> +#include <test/ut.h> + +/* Test that we can use the EFI_MEDIA uclass */ +static int dm_test_efi_media (struct unit_test_state *uts) +{ +struct udevice *dev; + +ut_assertok (uclass_first_device_err (UCLASS_EFI_MEDIA , &dev)); + +return 0 ; +} +DM_TEST (dm_test_efi_media,UT_TESTF_SCAN_PDATA |UT_TESTF_SCAN_FDT );
Re: [PATCH v2 15/39] efi: Add EFI uclass for media
Hi Heinrich,
Quoted 6 lines · Heinrich Schuchardt
Quoted 1 line · Simon Glass
At present
UCLASS_EFIis used to represent an EFI filesystem among other
Quoted 8 lines
UCLASS_EFIis for UEFI drivers that provide the driver binding protocol.Your new UCLASS seems to be for U-Boot drivers consuming the
EFI_BLOCK_IO_PROTOCOL.I agree that
lib/efi_driver/efi_block_device.cis to much intertwined with the UEFI subsystem to easily merge it with your requirement. But it at least can serve as a template.
It did.
Quoted 10 lines
Quoted 7 lines
things. The description of this uclass is "EFI managed devices" which is pretty vague. The only driver that uses this uclass is in fact not a real U-Boot driver, since its operations do not include a struct udevice.
Rather than mess with this, create a new
UCLASS_EFI_MEDIAuclass to handle EFI media such as disks. Make this the uclass to use for EFI media so thatIn UEFI speak MEDIA relates to:
Quoted 7 lines
* block devices
* firmware files
* Bluetooth
* REST services
* etcDo you only mean block devices?
Sort of...it is a parent of a block device, like MMC or SCSI.
Quoted 1 line
Why can't you use
UCLASS_BLKfor your new driver?
That is the child. I need something to be the parent so that it actually works.
Regards,
Simon
PATCH v2 16/39 efi: Add a media/block driver for EFI block devices
In reply to Simon GlassAdd a block driver which handles read/write for EFI block devices. This driver actually already exists ('efi_block') but is not really suitable for use as a real U-Boot driver:
- The operations do not provide a udevice
- The code is designed for running as part of EFI loader, so uses
EFI_PRINT() and EFI_CALL().
- The bind method probes the device, which is not permitted
- It uses 'EFI' as its parent device
The new driver is more 'normal', just requiring its platform data be set up in advance.
---
Changes in v2:
- Drop mention of partitions from the commit message
drivers/block/Kconfig | 10 ++++ drivers/block/Makefile | 1 + drivers/block/efi_blk.c | 115 ++++++++++++++++++++++++++++++++++++++++ include/efi.h | 11 ++++ 4 files changed, 137 insertions(+) create mode 100644 drivers/block/efi_blk.c
drivers/block/Kconfig +10 -0
@@ -82,6 +82,16 @@ config EFI_MEDIA_SANDBOX EFI_MEDIA uclass. It does not do anything useful, since sandbox does not actually support running on top of UEFI. +config EFI_MEDIA_BLK + bool "EFI media block driver" + depends on EFI_APP + default y + help + Enables a block driver for providing access to UEFI devices. This + allows use of block devices detected by the underlying UEFI + implementation. With this it is possible to use filesystems on these + devices, for example. + endif # EFI_MEDIA config IDE
drivers/block/Makefile +1 -0
@@ -17,3 +17,4 @@ obj-$(CONFIG_$(SPL_TPL_)BLOCK_CACHE) += blkcache.o obj-$(CONFIG_EFI_MEDIA) += efi-media-uclass.o obj-$(CONFIG_EFI_MEDIA_SANDBOX) += sb_efi_media.o +obj-$(CONFIG_EFI_MEDIA_BLK) += efi_blk.o
drivers/block/efi_blk.c new file +115 -0
115 lines not on this page · raw message
include/efi.h +11 -0
@@ -414,6 +414,17 @@ struct efi_priv {void *next_hdr; }; +/* +* EFI attributes of the udevice handled by efi_media driver +* +* @handle: handle of the controller on which this driver is installed +* @blkio: block io protocol proxied by this driver +*/ +struct efi_media_plat { +efi_handle_t handle; +struct efi_block_io *blkio; +}; +/* Base address of the EFI image */ extern char image_base[];
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 17/39 efi: Locate all block devices in the app
In reply to Simon GlassWhen starting the app, locate all block devices and make them available to U-Boot. This allows listing partitions and accessing files in filesystems.
EFI also has the concept of 'disks', meaning boot media. For now, this is not obviously useful in U-Boot, but add code to at least locate these. This can be expanded later as needed.
Series-changes; 2
- Store device path in struct efi_media_plat
- Don't export efi_bind_block()
- Only bind devices for media devices, not for partitions
- Show devices that are processed
- Update documentation
---
(no changes since v1)
doc/develop/uefi/u-boot_on_efi.rst | 4 +- include/efi.h | 6 +- include/efi_api.h | 15 ++ lib/efi/efi_app.c | 223 +++++++++++++++++++++++++++++ 4 files changed, 243 insertions(+), 5 deletions(-)
doc/develop/uefi/u-boot_on_efi.rst +1 -3
@@ -263,9 +263,7 @@ This work could be extended in a number of ways: - Figure out how to solve the interrupt problem -- Add more drivers to the application side (e.g. block devices, USB, - environment access). This would mostly be an academic exercise as a strong - use case is not readily apparent, but it might be fun. +- Add more drivers to the application side (e.g.USB, environment access). - Avoid turning off boot services in the stub. Instead allow U-Boot to make use of boot services in case it wants to. It is unclear what it might want
include/efi.h +4 -2
@@ -419,10 +419,12 @@ struct efi_priv {* * @handle: handle of the controller on which this driver is installed * @blkio: block io protocol proxied by this driver +* @device_path: EFI path to the device */ struct efi_media_plat { -efi_handle_t handle; -struct efi_block_io *blkio; +efi_handle_t handle; +struct efi_block_io *blkio; +struct efi_device_path *device_path; };/* Base address of the EFI image */
include/efi_api.h +15 -0
@@ -1994,4 +1994,19 @@ struct efi_firmware_management_protocol { const u16 *package_version_name); }; +#define EFI_DISK_IO_PROTOCOL_GUID \ + EFI_GUID(0xce345171, 0xba0b, 0x11d2, 0x8e, 0x4f, \ + 0x00, 0xa0, 0xc9, 0x69, 0x72, 0x3b) + +struct efi_disk { +u64 revision; + efi_status_t (EFIAPI *read_disk)(struct efi_disk *this, u32 media_id, + u64 offset, efi_uintn_t buffer_size, + void *buffer); + + efi_status_t (EFIAPI *write_disk)(struct efi_disk *this, u32 media_id, + u64 offset, efi_uintn_t buffer_size, + void *buffer); +}; +#endif
lib/efi/efi_app.c +223 -0
241 lines not on this page · raw message
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 18/39 patman: Use a ValueError exception if tools.Run() fails
In reply to Simon GlassThe Exception base class is a very vague and could be confusing to the test system. Use the more specific ValueError exception instead.
---
(no changes since v1)
tools/patman/tools.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
tools/patman/tools.py +1 -1
@@ -348,7 +348,7 @@ def Run(name, *args, **kwargs): result = command.RunPipe([all_args], capture=True, capture_stderr=True, env=env, raise_on_error=False, binary=binary) if result.return_code: - raise Exception("Error %d running '%s': %s" % + raise ValueError("Error %d running '%s': %s" % (result.return_code,' '.join(all_args), result.stderr)) return result.stdout
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 19/39 binman: Report an error if test files fail to compile
In reply to Simon GlassAt present any error from the 'make' command is silently swallowed by the test system. Fix this by showing it when detected.
---
(no changes since v1)
tools/binman/elf_test.py | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-)
tools/binman/elf_test.py +6 -2
@@ -70,8 +70,12 @@ def BuildElfTestFiles(target_dir): # correctly. So drop any make flags here. if 'MAKEFLAGS' in os.environ: del os.environ['MAKEFLAGS'] - tools.Run('make', '-C', target_dir, '-f', - os.path.join(testdir, 'Makefile'), 'SRC=%s/' % testdir) + try: + tools.Run('make', '-C', target_dir, '-f', + os.path.join(testdir, 'Makefile'), 'SRC=%s/' % testdir) + except ValueError as e: + # The test system seems to suppress this in a strange way + print(e) class TestElf(unittest.TestCase):
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 20/39 binman: Support reading the offset of an ELF-file symbol
In reply to Simon GlassBinman needs to be able to update the contents of an ELF file after it has been build. To support this, add a function to locate the position of a symbol's contents within the file.
Fix the comments on bss_data.c and Symbol while we are here.
---
(no changes since v1)
tools/binman/elf.py | 53 ++++++++++++++++++++++++++++++-- tools/binman/elf_test.py | 37 ++++++++++++++++++++++ tools/binman/test/Makefile | 5 ++- tools/binman/test/bss_data.c | 2 +- tools/binman/test/embed_data.c | 16 ++++++++++ tools/binman/test/embed_data.lds | 23 ++++++++++++++ 6 files changed, 131 insertions(+), 5 deletions(-) create mode 100644 tools/binman/test/embed_data.c create mode 100644 tools/binman/test/embed_data.lds
tools/binman/elf.py +50 -3
@@ -24,7 +24,14 @@ try: except: # pragma: no cover ELF_TOOLS = False -Symbol = namedtuple('Symbol', ['section', 'address', 'size', 'weak']) +# Information about an EFL symbol: +# section (str): Name of the section containing this symbol +# address (int): Address of the symbol (its value) +# size (int): Size of the symbol in bytes +# weak (bool): True if the symbol is weak +# offset (int or None): Offset of the symbol's data in the ELF file, or None if +# not known +Symbol = namedtuple('Symbol', ['section', 'address', 'size', 'weak', 'offset']) # Information about an ELF file: # data: Extracted program contents of ELF file (this would be loaded by an
@@ -71,8 +78,48 @@ def GetSymbols(fname, patterns): section, size = parts[:2] if len(parts) > 2: name = parts[2] if parts[2] != '.hidden' else parts[3] - syms[name] = Symbol(section, int(value, 16), int(size,16), - flags[1] == 'w') + syms[name] = Symbol(section, int(value, 16), int(size, 16), + flags[1] == 'w', None) + + # Sort dict by address + return OrderedDict(sorted(syms.items(), key=lambda x: x[1].address)) + +def GetSymbolFileOffset(fname, patterns): + """Get the symbols from an ELF file + + Args: + fname: Filename of the ELF file to read + patterns: List of regex patterns to search for, each a string + + Returns: + None, if the file does not exist, or Dict: + key: Name of symbol + value: Hex value of symbol + """ + def _GetFileOffset(elf, addr): + for seg in elf.iter_segments(): + seg_end = seg['p_vaddr'] + seg['p_filesz'] + if seg.header['p_type'] == 'PT_LOAD': + if addr >= seg['p_vaddr'] and addr < seg_end: + return addr - seg['p_vaddr'] + seg['p_offset'] + + if not ELF_TOOLS: + raise ValueError('Python elftools package is not available') + + syms = {} + with open(fname, 'rb') as fd: + elf = ELFFile(fd) + + re_syms = re.compile('|'.join(patterns)) + for section in elf.iter_sections(): + if isinstance(section, SymbolTableSection): + for symbol in section.iter_symbols(): + if not re_syms or re_syms.search(symbol.name): + addr = symbol.entry['st_value'] + syms[symbol.name] = Symbol( + section.name, addr, symbol.entry['st_size'], + symbol.entry['st_info']['bind'] == 'STB_WEAK', + _GetFileOffset(elf, addr)) # Sort dict by address return OrderedDict(sorted(syms.items(), key=lambda x: x[1].address))
tools/binman/elf_test.py +37 -0
@@ -6,6 +6,7 @@ import os import shutil +import struct import sys import tempfile import unittest
@@ -221,6 +222,42 @@ class TestElf(unittest.TestCase): elf.DecodeElf(data, load + 2)) shutil.rmtree(outdir) + def testEmbedData(self): + """Test for the GetSymbolFileOffset() function""" + if not elf.ELF_TOOLS: + self.skipTest('Python elftools not available') + + fname = self.ElfTestFile('embed_data') + offset = elf.GetSymbolFileOffset(fname, ['embed_start', 'embed_end']) + start = offset['embed_start'].offset + end = offset['embed_end'].offset + data = tools.ReadFile(fname) + embed_data = data[start:end] + expect = struct.pack('<III', 0x1234, 0x5678, 0) + self.assertEqual(expect, embed_data) + + def testEmbedFail(self): + """Test calling GetSymbolFileOffset() without elftools""" + try: + old_val = elf.ELF_TOOLS + elf.ELF_TOOLS = False + fname = self.ElfTestFile('embed_data') + with self.assertRaises(ValueError) as e: + elf.GetSymbolFileOffset(fname, ['embed_start', 'embed_end']) + self.assertIn('Python elftools package is not available', + str(e.exception)) + finally: + elf.ELF_TOOLS = old_val + + def testEmbedDataNoSym(self): + """Test for GetSymbolFileOffset() getting no symbols""" + if not elf.ELF_TOOLS: + self.skipTest('Python elftools not available') + + fname = self.ElfTestFile('embed_data') + offset = elf.GetSymbolFileOffset(fname, ['missing_sym']) + self.assertEqual({}, offset) + if __name__ == '__main__': unittest.main()
tools/binman/test/Makefile +4 -1
@@ -31,7 +31,7 @@ LDS_BINMAN_X86 := -T $(SRC)u_boot_binman_syms_x86.lds TARGETS = u_boot_ucode_ptr u_boot_no_ucode_ptr bss_data \ u_boot_binman_syms u_boot_binman_syms.bin u_boot_binman_syms_bad \ - u_boot_binman_syms_size u_boot_binman_syms_x86 + u_boot_binman_syms_size u_boot_binman_syms_x86 embed_data all: $(TARGETS)
@@ -44,6 +44,9 @@ u_boot_ucode_ptr: u_boot_ucode_ptr.c bss_data: CFLAGS += $(SRC)bss_data.lds bss_data: bss_data.c +embed_data: CFLAGS += $(SRC)embed_data.lds +embed_data: embed_data.c + u_boot_binman_syms.bin: u_boot_binman_syms $(OBJCOPY) -O binary $< -R .note.gnu.build-id $@
tools/binman/test/bss_data.c +1 -1
@@ -2,7 +2,7 @@/* * Copyright (c) 2016 Google, Inc * -* Simple program to create a _dt_ucode_base_size symbol which can be read +* Simple program to create a bss_data region so the symbol can be read * by binutils. This is used by binman tests. */
tools/binman/test/embed_data.c new file +16 -0
@@ -0,0 +1,16 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* Copyright 2021 Google LLC +* +* Simple program including some embedded data that can be accessed by binman. +* This is used by binman tests. +*/ + +int first[10 ] = {1 }; +int embed[3] __attribute__((section(".embed"))) = {0x1234, 0x5678}; +int second[10 ] = {1 }; + +int main (void ) +{ +return 0 ; +}
tools/binman/test/embed_data.lds new file +23 -0
@@ -0,0 +1,23 @@ +/* SPDX-License-Identifier: GPL-2.0+ */ +/* + * Copyright 2021 Google LLC + */ + +OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386") +OUTPUT_ARCH(i386) +ENTRY(_start) + +SECTIONS +{ + _start = .; + __data_start = .; + .data : + { + . = ALIGN(32); + embed_start = .; + *(.embed*) + embed_end = .; + . = ALIGN(32); + *(.data*) + } +}
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 21/39 binman: Allow timeout to occur in the image or its section
In reply to Simon GlassAt present testThreadTimeout() assumes that the expected timeout happens first when building the section, but it can just as easily happen at the top-level image. Update the test to cope with both.
---
(no changes since v1)
tools/binman/ftest.py | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
tools/binman/ftest.py +1 -2
@@ -4565,8 +4565,7 @@ class TestFunctional(unittest.TestCase): with self.assertRaises(ValueError) as e: self._DoTestFile('202_section_timeout.dts', test_section_timeout=True) - self.assertIn("Node '/binman/section@0': Timed out obtaining contents", - str(e.exception)) + self.assertIn("Timed out obtaining contents", str(e.exception)) def testTiming(self): """Test output of timing information"""
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 22/39 binman: Tidy up comments on _DoTestFile()
In reply to Simon GlassThe comment for this function is missing an argument and the return value. Fix it.
---
(no changes since v1)
tools/binman/ftest.py | 5 +++++ 1 file changed, 5 insertions(+)
tools/binman/ftest.py +5 -0
@@ -334,6 +334,11 @@ class TestFunctional(unittest.TestCase): extra_indirs: Extra input directories to add using -I threads: Number of threads to use (None for default, 0 for single-threaded) + test_section_timeout: True to force the first time to timeout, as + used in testThreadTimeout() + + Returns: + int return code, 0 on success """ args = [] if debug:
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 23/39 binman: Support updating the dtb in an ELF file
In reply to Simon GlassWIth EFI we must embed the devicetree in an ELF image so that it is loaded as part of the executable file. We want it to include the binman definition in there also, which in some cases cannot be created until the ELF (u-boot) is built. Add an option to binman to support writing the updated dtb to the ELF file u-boot.out
This is useful with the EFI app, which is always packaged as an ELF file.
---
(no changes since v1)
tools/binman/binman.rst | 36 ++++++++++ tools/binman/cmdline.py | 2 + tools/binman/control.py | 11 +++ tools/binman/elf.py | 21 ++++++ tools/binman/ftest.py | 83 +++++++++++++++++++++- tools/binman/test/Makefile | 10 ++- tools/binman/test/u_boot_binman_embed.c | 13 ++++ tools/binman/test/u_boot_binman_embed.lds | 29 ++++++++ tools/binman/test/u_boot_binman_embed_sm.c | 13 ++++ 9 files changed, 216 insertions(+), 2 deletions(-) create mode 100644 tools/binman/test/u_boot_binman_embed.c create mode 100644 tools/binman/test/u_boot_binman_embed.lds create mode 100644 tools/binman/test/u_boot_binman_embed_sm.c
tools/binman/binman.rst +36 -0
@@ -822,6 +822,42 @@ the 'warning' line in scripts/Makefile.lib to see what it has found:: # u_boot_dtsi_options_debug = $(u_boot_dtsi_options_raw) +Updating an ELF file +==================== + +For the EFI app, where U-Boot is loaded from UEFI and runs as an app, there is +no way to update the devicetree after U-Boot is built. Normally this works by +creating a new u-boot.dtb.out with he updated devicetree, which is automatically +built into the output image. With ELF this is not possible since the ELF is +not part of an image, just a stand-along file. We must create an updated ELF +file with the new devicetree. + +This is handled by the --update-fdt-in-elf option. It takes four arguments, +separated by comma: + + infile - filename of input ELF file, e.g. 'u-boot's + outfile - filename of output ELF file, e.g. 'u-boot.out' + begin_sym - symbol at the start of the embedded devicetree, e.g. + '__dtb_dt_begin' + end_sym - symbol at the start of the embedded devicetree, e.g. + '__dtb_dt_end' + +When this flag is used, U-Boot does all the normal packaging, but as an +additional step, it creates a new ELF file with the new devicetree embedded in +it. + +If logging is enabled you will see a message like this:: + + Updating file 'u-boot' with data length 0x400a (16394) between symbols + '__dtb_dt_begin' and '__dtb_dt_end' + +There must be enough space for the updated devicetree. If not, an error like +the following is produced:: + + ValueError: Not enough space in 'u-boot' for data length 0x400a (16394); + size is 0x1744 (5956) + + Entry Documentation ===================
tools/binman/cmdline.py +2 -0
@@ -71,6 +71,8 @@ controlled by a description in the board device tree.''' 'given') build_parser.add_argument('-u', '--update-fdt', action='store_true', default=False, help='Update the binman node with offset/size info') + build_parser.add_argument('--update-fdt-in-elf', type=str, + help='Update an ELF file with the output dtb: infile,outfile,begin_sym,end_sym') entry_parser = subparsers.add_parser('entry-docs', help='Write out entry documentation (see entries.rst)')
tools/binman/control.py +11 -0
@@ -598,6 +598,13 @@ def Binman(args): tools.FinaliseOutputDir() return 0 + elf_params = None + if args.update_fdt_in_elf: + elf_params = args.update_fdt_in_elf.split(',') + if len(elf_params) != 4: + raise ValueError('Invalid args %s to --update-fdt-in-elf: expected infile,outfile,begin_sym,end_sym' % + elf_params) + # Try to figure out which device tree contains our image description if args.dt: dtb_fname = args.dt
@@ -644,6 +651,10 @@ def Binman(args): for dtb_item in state.GetAllFdts(): tools.WriteFile(dtb_item._fname, dtb_item.GetContents()) + if elf_params: + data = state.GetFdtForEtype('u-boot-dtb').GetContents() + elf.UpdateFile(*elf_params, data) + if missing: tout.Warning("\nSome images are invalid")
tools/binman/elf.py +21 -0
@@ -348,3 +348,24 @@ def DecodeElf(data, location): segment.data()[offset:]) return ElfInfo(output, data_start, elf.header['e_entry'] + virt_to_phys, mem_end - data_start) + +def UpdateFile(infile, outfile, start_sym, end_sym, insert): + tout.Notice("Creating file '%s' with data length %#x (%d) between symbols '%s' and '%s'" % + (outfile, len(insert), len(insert), start_sym, end_sym)) + syms = GetSymbolFileOffset(infile, [start_sym, end_sym]) + if len(syms) != 2: + raise ValueError("Expected two symbols '%s' and '%s': got %d: %s" % + (start_sym, end_sym, len(syms), + ','.join(syms.keys()))) + + size = syms[end_sym].offset - syms[start_sym].offset + if len(insert) > size: + raise ValueError("Not enough space in '%s' for data length %#x (%d); size is %#x (%d)" % + (infile, len(insert), len(insert), size, size)) + + data = tools.ReadFile(infile) + newdata = data[:syms[start_sym].offset] + newdata += insert + tools.GetBytes(0, size - len(insert)) + newdata += data[syms[end_sym].offset:] + tools.WriteFile(outfile, newdata) + tout.Info('Written to offset %#x' % syms[start_sym].offset)
tools/binman/ftest.py +82 -1
107 lines not on this page · raw message
tools/binman/test/Makefile +9 -1
@@ -28,10 +28,12 @@ LDS_UCODE := -T $(SRC)u_boot_ucode_ptr.lds LDS_BINMAN := -T $(SRC)u_boot_binman_syms.lds LDS_BINMAN_BAD := -T $(SRC)u_boot_binman_syms_bad.lds LDS_BINMAN_X86 := -T $(SRC)u_boot_binman_syms_x86.lds +LDS_BINMAN_EMBED := -T $(SRC)u_boot_binman_embed.lds TARGETS = u_boot_ucode_ptr u_boot_no_ucode_ptr bss_data \ u_boot_binman_syms u_boot_binman_syms.bin u_boot_binman_syms_bad \ - u_boot_binman_syms_size u_boot_binman_syms_x86 embed_data + u_boot_binman_syms_size u_boot_binman_syms_x86 embed_data \ + u_boot_binman_embed u_boot_binman_embed_sm all: $(TARGETS)
@@ -62,6 +64,12 @@ u_boot_binman_syms_bad: u_boot_binman_syms_bad.c u_boot_binman_syms_size: CFLAGS += $(LDS_BINMAN) u_boot_binman_syms_size: u_boot_binman_syms_size.c +u_boot_binman_embed: CFLAGS += $(LDS_BINMAN_EMBED) +u_boot_binman_embed: u_boot_binman_embed.c + +u_boot_binman_embed_sm: CFLAGS += $(LDS_BINMAN_EMBED) +u_boot_binman_embed_sm: u_boot_binman_embed_sm.c + clean: rm -f $(TARGETS)
tools/binman/test/u_boot_binman_embed.c new file +13 -0
@@ -0,0 +1,13 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* Copyright 2021 Google LLC +* +* Simple program to embed a devicetree. This is used by binman tests. +*/ + +int __attribute__((section (".mydtb" ))) dtb_data[4096 ]; + +int main (void ) +{ +return 0 ; +}
tools/binman/test/u_boot_binman_embed.lds new file +29 -0
@@ -0,0 +1,29 @@ +/* SPDX-License-Identifier: GPL-2.0+ */ +/* + * Copyright (c) 2016 Google, Inc + */ + +OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386") +OUTPUT_ARCH(i386) +ENTRY(_start) + +SECTIONS +{ + . = 0x00000000; + _start = .; + + . = ALIGN(4); + .text : + { + *(.text*) + } + + . = ALIGN(4); + .data : { + dtb_embed_begin = .; + KEEP(*(.mydtb)); + dtb_embed_end = .; + } + .interp : { *(.interp*) } + +}
tools/binman/test/u_boot_binman_embed_sm.c new file +13 -0
@@ -0,0 +1,13 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* Copyright 2021 Google LLC +* +* Simple program to embed a devicetree. This is used by binman tests. +*/ + +int __attribute__((section (".mydtb" ))) dtb_data[16 ]; + +int main (void ) +{ +return 0 ; +}
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 24/39 efi: serial: Support arrow keys
In reply to Simon GlassAt present only the backspace key is supported in U-Boot, when running as an EFI app. Add support for arrows, home and end as well, to make the CLI more friendly.
---
(no changes since v1)
drivers/serial/serial_efi.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-)
drivers/serial/serial_efi.c +9 -2
@@ -24,6 +24,9 @@ struct serial_efi_priv {bool have_key; }; +/* Convert a lower-case character to its ctrl-char equivalent */ +#define CTL_CH (c) ((c) - 'a' + 1) +int serial_efi_setbrg (struct udevice *dev,int baudrate) {return 0 ;
@@ -49,6 +52,7 @@ static int serial_efi_get_key(struct serial_efi_priv *priv)static int serial_efi_getc (struct udevice *dev) {struct serial_efi_priv *priv =dev_get_priv (dev); +char conv_scan[10 ] = {0 ,'p' ,'n' ,'f' ,'b' ,'a' ,'e' ,0 ,8 };int ret, ch; ret =serial_efi_get_key (priv);
@@ -63,8 +67,11 @@ static int serial_efi_getc(struct udevice *dev)* key scan code of 8. Handle this so that backspace works correctly * in the U-Boot command line. */ -if (!ch && priv->key.scan_code ==8 ) - ch =8 ; +if (!ch && priv->key.scan_code <sizeof (conv_scan)) { + ch = conv_scan[priv->key.scan_code]; +if (ch >='a' ) + ch -='a' -1 ; + }debug (" [%x %x %x] " , ch, priv->key.unicode_char, priv->key.scan_code);return ch;
Signature · 1 line
2.33.0.685.g46640cef36-goog
Re: [PATCH v2 24/39] efi: serial: Support arrow keys
Quoted 35 lines · Simon Glass
At present only the backspace key is supported in U-Boot, when running as an EFI app. Add support for arrows, home and end as well, to make the CLI more friendly.
Signed-off-by Simon Glass <[email protected]>---
(no changes since v1)
drivers/serial/serial_efi.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-)drivers/serial/serial_efi.c +4
-0@@ -24,6 +24,9 @@ struct serial_efi_priv {bool have_key; }; +/* Convert a lower-case character to its ctrl-char equivalent */ +#define CTL_CH (c) ((c) - 'a' + 1) +int serial_efi_setbrg (struct udevice *dev,int baudrate) {return 0 ;@@ -49,6 +52,7 @@ static int serial_efi_get_key(struct serial_efi_priv *priv)static int serial_efi_getc (struct udevice *dev) {struct serial_efi_priv *priv =dev_get_priv (dev); +char conv_scan[10 ] = {0 ,'p' ,'n' ,'f' ,'b' ,'a' ,'e' ,0 ,8 };int ret, ch; ret =serial_efi_get_key (priv);
drivers/serial/serial_efi.c static int serial_efi_getc(struct udevice *dev) · line 73
Quoted 9 lines
drivers/serial/serial_efi.c +4
-2@@ -63,8 +67,11 @@ static int serial_efi_getc(struct udevice *dev)* key scan code of 8. Handle this so that backspace works correctly * in the U-Boot command line. */ -if (!ch && priv->key.scan_code ==8 ) - ch =8 ; +if (!ch && priv->key.scan_code <sizeof (conv_scan)) + ch = conv_scan[priv->key.scan_code]; +if (ch >='a' ) + ch -='a' -1 ;
All 23 EFI Scan Codes should be supported. Can't we use something like kbd_to_ansi364[] from drivers/input/input.c to create the appropriate escape sequences?
Best regards
Heinrich
Quoted 5 lines
drivers/serial/serial_efi.c +1
-0+ }debug (" [%x %x %x] " , ch, priv->key.unicode_char, priv->key.scan_code);return ch;
Re: [PATCH v2 24/39] efi: serial: Support arrow keys
Hi Heinrich,
Quoted 35 lines · Heinrich Schuchardt
Quoted 33 lines · Simon Glass
At present only the backspace key is supported in U-Boot, when running as an EFI app. Add support for arrows, home and end as well, to make the CLI more friendly.
Signed-off-by Simon Glass <[email protected]>---
(no changes since v1)
drivers/serial/serial_efi.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-)drivers/serial/serial_efi.c +4
-0@@ -24,6 +24,9 @@ struct serial_efi_priv {bool have_key; }; +/* Convert a lower-case character to its ctrl-char equivalent */ +#define CTL_CH (c) ((c) - 'a' + 1) +int serial_efi_setbrg (struct udevice *dev,int baudrate) {return 0 ;@@ -49,6 +52,7 @@ static int serial_efi_get_key(struct serial_efi_priv *priv)static int serial_efi_getc (struct udevice *dev) {struct serial_efi_priv *priv =dev_get_priv (dev); +char conv_scan[10 ] = {0 ,'p' ,'n' ,'f' ,'b' ,'a' ,'e' ,0 ,8 };int ret, ch;
Quoted 15 lines
drivers/serial/serial_efi.c static int serial_efi_getc(struct udevice *dev) · line 73
Quoted 11 lines
drivers/serial/serial_efi.c +4
-2ret =serial_efi_get_key (priv);@@ -63,8 +67,11 @@ static int serial_efi_getc(struct udevice *dev)* key scan code of 8. Handle this so that backspace works correctly * in the U-Boot command line. */ -if (!ch && priv->key.scan_code ==8 ) - ch =8 ; +if (!ch && priv->key.scan_code <sizeof (conv_scan)) + ch = conv_scan[priv->key.scan_code]; +if (ch >='a' ) + ch -='a' -1 ;All 23 EFI Scan Codes should be supported. Can't we use something like kbd_to_ansi364[] from
drivers/input/input.cto create the appropriate escape sequences?
I think you are referring to this:
Table 107. EFI Scan Codes for EFI_SIMPLE_TEXT_INPUT_PROTOCOL EFI Scan Code Description0x00 Null scan code.0x01 Move cursor up 1 row.0x02 Move cursor down 1 row.0x03 Move cursor right 1 column.0x04 Move cursor left 1 column.0x05 Home.0x06 End.0x07 Insert.0x08 Delete.0x09 Page Up.0x0a Page Down.0x0b Function 1.0x0c Function 2.0x0d Function 3.0x0e Function 4.0x0f Function 5.0x10 Function 6.0x11 Function 7.0x12 Function 8.0x13 Function 9.0x14 Function 10.
Which ones (other than what I have added) do you think U-Boot supports? I cannot find any.
Regards,
Simon
PATCH v2 25/39 bloblist: Support allocating the bloblist
In reply to Simon GlassTypically the bloblist is positioned at a fixed address in memory until relocation. This is convenient when it is set up in SPL or before relocation.
But for EFI we want to set it up only when U-Boot proper is running. Add a way to allocate it using malloc() and update the documentation to cover this aspect of bloblist.
Note there are no tests of this feature at present, nor any direct testing of bloblist_init().
This can be added, e.g. by making this option controllable at runtime.
---
(no changes since v1)
common/Kconfig | 15 +++++++++++++-- common/bloblist.c | 16 ++++++++++++++-- common/board_f.c | 8 +++++++- doc/develop/bloblist.rst | 16 ++++++++++++++++ 4 files changed, 50 insertions(+), 5 deletions(-)
common/Kconfig +13 -2
@@ -719,6 +719,8 @@ config TPL_BLOBLIST This enables a bloblist in TPL. The bloblist is set up in TPL and passed to SPL and U-Boot proper. +if BLOBLIST + config BLOBLIST_SIZE hex "Size of bloblist" depends on BLOBLIST
@@ -729,17 +731,24 @@ config BLOBLIST_SIZE is set up in the first part of U-Boot to run (TPL, SPL or U-Boot proper), and this sane bloblist is used for subsequent stages. +config BLOBLIST_ALLOC + bool "Allocate bloblist" + help + Allocate the bloblist using malloc(). This avoids the need to + specify a fixed address on systems where this is unknown or can + change at runtime. + config BLOBLIST_ADDR hex "Address of bloblist" - depends on BLOBLIST default 0xc000 if SANDBOX help Sets the address of the bloblist, set up by the first part of U-Boot which runs. Subsequent U-Boot stages typically use the same address. + This is not used if BLOBLIST_ALLOC is selected. + config BLOBLIST_SIZE_RELOC hex "Size of bloblist after relocation" - depends on BLOBLIST default BLOBLIST_SIZE help Sets the size of the bloblist in bytes after relocation. Since U-Boot
@@ -747,6 +756,8 @@ config BLOBLIST_SIZE_RELOC size than the one set up by SPL. This bloblist is set up during the relocation process. +endif # BLOBLIST + endmenu source "common/spl/Kconfig"
common/bloblist.c +14 -2
@@ -7,6 +7,7 @@#include <common.h> #include <bloblist.h> #include <log.h> +#include <malloc.h> #include <mapmem.h> #include <spl.h> #include <asm/global_data.h>
@@ -416,10 +417,21 @@ int bloblist_init(void) ret =bloblist_check (CONFIG_BLOBLIST_ADDR ,CONFIG_BLOBLIST_SIZE );if (ret) { +ulong addr; +log (LOGC_BLOBLIST , expected ?LOGL_WARNING :LOGL_DEBUG ,"Existing bloblist not found: creating new bloblist\n" ); - ret =bloblist_new (CONFIG_BLOBLIST_ADDR ,CONFIG_BLOBLIST_SIZE , -0 ); +if (IS_ENABLED (CONFIG_BLOBLIST_ALLOC )) { +void *ptr =memalign (BLOBLIST_ALIGN , +CONFIG_BLOBLIST_SIZE ); + +if (!ptr) +return log_msg_ret ("alloc" , -ENOMEM ); + addr =map_to_sysmem (ptr); + }else { + addr =CONFIG_BLOBLIST_ADDR ; + } + ret =bloblist_new (addr,CONFIG_BLOBLIST_SIZE ,0 ); }else {log (LOGC_BLOBLIST ,LOGL_DEBUG ,"Found existing bloblist\n" ); }
common/board_f.c +7 -1
@@ -655,8 +655,14 @@ static int reloc_bootstage(void)static int reloc_bloblist (void ) {#ifdef CONFIG_BLOBLIST -if (gd->flags &GD_FLG_SKIP_RELOC ) +/* +* Relocate only if we are supposed to send it +*/ +if ((gd->flags &GD_FLG_SKIP_RELOC ) && +CONFIG_BLOBLIST_SIZE ==CONFIG_BLOBLIST_SIZE_RELOC ) { +debug ("Not relocating bloblist\n" );return 0 ; + }if (gd->new_bloblist) {int size =CONFIG_BLOBLIST_SIZE ;
doc/develop/bloblist.rst +16 -0
@@ -59,6 +59,22 @@ Bloblist provides a fairly simple API which allows blobs to be created and found. All access is via the blob's tag. Blob records are zeroed when added. +Placing the bloblist +-------------------- + +The bloblist is typically positioned at a fixed address by TPL, or SPL. This +is controlled by `CONFIG_BLOBLIST_ADDR`. But in some cases it is preferable to +allocate the bloblist in the malloc() space. Use the `CONFIG_BLOBLIST_ALLOC` +option to enable this. + +The bloblist is automatically relocated as part of U-Boot relocation. Sometimes +it is useful to expand the bloblist in U-Boot proper, since it may want to add +information for use by Linux. Note that this does not mean that Linux needs to +know anything about the bloblist format, just that it is convenient to use +bloblist to place things contiguously in memory. Set +`CONFIG_BLOBLIST_SIZE_RELOC` to define the expanded size, if needed. + + Finishing the bloblist ----------------------
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 26/39 x86: Allow booting a kernel from the EFI app
In reply to Simon GlassAt present this is disabled, but it should work so long as the kernel does not need EFI services. Enable it and add a note about remaining work.
---
Changes in v2:
- Update documentation
arch/x86/lib/bootm.c | 11 +++++++---- doc/develop/uefi/u-boot_on_efi.rst | 2 +- 2 files changed, 8 insertions(+), 5 deletions(-)
arch/x86/lib/bootm.c +7 -4
@@ -179,10 +179,14 @@ int boot_linux_kernel(ulong setup_base, ulong load_address, bool image_64bit)* U-Boot is setting them up that way for itself in * arch/i386/cpu/cpu.c. * -* Note that we cannot currently boot a kernel while running as -* an EFI application. Please use the payload option for that. +* Note: this is incomplete for EFI kernels! +* +* This can boot a kernel while running as an EFI application, +* but if the kernel requires EFI support then that support needs +* to be enabled first (see EFI_LOADER). Also the EFI information +* must enabled with setup_efi_info(). See setup_zimage() for +* how this is done with the stub. */ -#ifndef CONFIG_EFI_APP __asm__ __volatile__ ("movl $0, %%ebp\n" "cli\n"
@@ -191,7 +195,6 @@ int boot_linux_kernel(ulong setup_base, ulong load_address, bool image_64bit) [boot_params] "S"(setup_base),"b" (0 ),"D" (0 ) ); -#endif }/* We can't get to here */
doc/develop/uefi/u-boot_on_efi.rst +1 -1
@@ -267,7 +267,7 @@ This work could be extended in a number of ways: - Avoid turning off boot services in the stub. Instead allow U-Boot to make use of boot services in case it wants to. It is unclear what it might want - though. + though. It is better to use the app. Where is the code? ------------------
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 27/39 x86: Don't process the kernel command line unless enabled
In reply to Simon GlassIf the 'bootm' command is not enabled then this code is not available and this causes a link error. Fix it.
Note that for the EFI app, there is no indication of missing code. It just hangs!
---
(no changes since v1)
arch/x86/lib/zimage.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-)
arch/x86/lib/zimage.c +8 -5
@@ -365,11 +365,14 @@ int setup_zimage(struct boot_params *setup_base, char *cmd_line, int auto_boot,strcpy (cmd_line, (char *)cmdline_force); elsebuild_command_line (cmd_line ,auto_boot ); - ret =bootm_process_cmdline (cmd_line, max_size,BOOTM_CL_ALL ); -if (ret) { -printf ("Cmdline setup failed (max_size=%x, bootproto=%x, err=%d)\n" , - max_size, bootproto, ret); -return ret; +if (IS_ENABLED (CONFIG_CMD_BOOTM )) { + ret =bootm_process_cmdline (cmd_line, max_size, +BOOTM_CL_ALL ); +if (ret) { +printf ("Cmdline setup failed (max_size=%x, bootproto=%x, err=%d)\n" , + max_size, bootproto, ret); +return ret; + } }printf ("Kernel command line: \"" );puts (cmd_line);
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 28/39 x86: efi: Add room for the binman definition in the dtb
In reply to Simon GlassAt present only 4KB of spare space is left in the DTB when building the EFI app. Increase this to 32KB so there is plenty of space to insert the binman definition. This cannot be expanded later (as with OF_SEPARATE) because the ELF image has already been built.
---
(no changes since v1)
arch/x86/dts/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/dts/Makefile +1 -1
@@ -24,7 +24,7 @@ dtb-y += bayleybay.dtb \ targets += $(dtb-y) -DTC_FLAGS += -R 4 -p 0x1000 +DTC_FLAGS += -R 4 -p $(if $(CONFIG_EFI_APP),0x8000,0x1000) PHONY += dtbs dtbs: $(addprefix $(obj)/, $(dtb-y))
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 29/39 efi: Add comments to struct efi_priv
In reply to Simon GlassThis structure is uncommented. Fix it.
---
(no changes since v1)
include/efi.h | 26 +++++++++++++++++++++++++- 1 file changed, 25 insertions(+), 1 deletion(-)
include/efi.h +25 -1
@@ -400,15 +400,39 @@ static inline struct efi_mem_desc *efi_get_next_mem_desc(return (struct efi_mem_desc *)((ulong )desc + map->desc_size); } +/** +* struct efi_priv - Information about the environment provided by EFI +* +* @parent_image: image passed into the EFI app or stub +* @sys_table: Pointer to system table +* @boot: Pointer to boot-services table +* @run: Pointer to runtime-services table +* +* Used by app only: +* @use_pool_for_malloc: true if all allocation should go through the EFI 'pool' +* methods allocate_pool() and free_pool(); false to use 'pages' methods +* allocate_pages() and free_pages() +* @ram_base: Base address of RAM (size CONFIG_EFI_RAM_SIZE) +* @image_data_type: Type of the loaded image (e.g. EFI_LOADER_CODE) +* +* Used by stub only: +* @info: Header of the info list, holding info collected by the stub and passed +* to U-Boot +* @info_size: Size of the info list, in bytes from @info +* @next_hdr: Pointer to where to put the next header when adding to the list +*/ struct efi_priv {efi_handle_t parent_image; -struct efi_device_path *device_path;struct efi_system_table *sys_table;struct efi_boot_services *boot;struct efi_runtime_services *run; + +/* app: */ bool use_pool_for_malloc;unsigned long ram_base;unsigned int image_data_type; + +/* stub: */ struct efi_info_hdr *info;unsigned int info_size;void *next_hdr;
Signature · 1 line
2.33.0.685.g46640cef36-goog
Re: [PATCH v2 29/39] efi: Add comments to struct efi_priv
Quoted 1 line · Simon Glass
This structure is uncommented. Fix it.
The commit message is incomplete. You are deleting field device path.
Quoted 16 lines
Signed-off-by Simon Glass <[email protected]>---
(no changes since v1)
include/efi.h | 26 +++++++++++++++++++++++++- 1 file changed, 25 insertions(+), 1 deletion(-)include/efi.h +0
-0@@ -400,15 +400,39 @@ static inline struct efi_mem_desc *efi_get_next_mem_desc(return (struct efi_mem_desc *)((ulong )desc + map->desc_size); }
include/efi.h static inline struct efi_mem_desc *efi_get_next_mem_desc( · line 411
Quoted 10 lines
include/efi.h +9
-0+/** +* struct efi_priv - Information about the environment provided by EFI +* +* @parent_image: image passed into the EFI app or stub +* @sys_table: Pointer to system table +* @boot: Pointer to boot-services table +* @run: Pointer to runtime-services table +* +* Used by app only:
Sphinx will render this line *after* all parameters. Please, remove it.
<cite>
...info_size
Size of the info list, in bytes from info
next_hdr
Pointer to where to put the next header when adding to the list
Description
-----------
Used by app only: Used by stub only:
struct efi_system_table *efi_get_sys_table(void)
</cite>
include/efi.h static inline struct efi_mem_desc *efi_get_next_mem_desc( · line 418
Quoted 7 lines
include/efi.h +7
-0+* @use_pool_for_malloc: true if all allocation should go through the EFI 'pool' +* methods allocate_pool() and free_pool(); false to use 'pages' methods +* allocate_pages() and free_pages() +* @ram_base: Base address of RAM (size CONFIG_EFI_RAM_SIZE) +* @image_data_type: Type of the loaded image (e.g. EFI_LOADER_CODE) +* +* Used by stub only:
Remove this line.
include/efi.h static inline struct efi_mem_desc *efi_get_next_mem_desc( · line 405
Quoted 8 lines
include/efi.h +5
-1+* @info: Header of the info list, holding info collected by the stub and passed +* to U-Boot +* @info_size: Size of the info list, in bytes from @info +* @next_hdr: Pointer to where to put the next header when adding to the list +*/ struct efi_priv {efi_handle_t parent_image; -struct efi_device_path *device_path;
This change seems to be unrelated.
Best regards
Heinrich
Quoted 14 lines
include/efi.h +4
-0struct efi_system_table *sys_table;struct efi_boot_services *boot;struct efi_runtime_services *run; + +/* app: */ bool use_pool_for_malloc;unsigned long ram_base;unsigned int image_data_type; + +/* stub: */ struct efi_info_hdr *info;unsigned int info_size;void *next_hdr;
PATCH v2 30/39 efi: Fix ll_boot_init() operation with the app
In reply to Simon GlassThis should return false when the EFI app is running, since UEFI has done the required low-level init. Fix it.
---
(no changes since v1)
include/init.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
include/init.h +1 -1
@@ -15,7 +15,7 @@#include <linux/types.h> /* Avoid using CONFIG_EFI_STUB directly as we may boot from other loaders */ -#ifdef CONFIG_EFI_STUB +#ifdef CONFIG_EFI #define ll_boot_init () false#else #include <asm/global_data.h>
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 32/39 efi: Share struct efi_priv between the app and stub code
In reply to Simon GlassAt present each of these has its own static variable and helper functions. Move them into a shared file.
---
(no changes since v1)
include/efi.h | 21 +++++++++++++++++++++ lib/efi/efi.c | 29 +++++++++++++++++++++++++++++ lib/efi/efi_app.c | 21 ++------------------- lib/efi/efi_stub.c | 7 ++++--- 4 files changed, 56 insertions(+), 22 deletions(-)
include/efi.h +21 -0
@@ -476,6 +476,27 @@ extern char _binary_u_boot_bin_start[], _binary_u_boot_bin_end[]; EFI_VARIABLE_TIME_BASED_AUTHENTICATED_WRITE_ACCESS | \ EFI_VARIABLE_APPEND_WRITE) +/** +* efi_get_priv() - Get access to the EFI-private information +* +* This struct it used by both the stub and the app to record things about the +* EFI environment. It is not available in U-Boot proper after the stub has +* jumped there. Use efi_info_get() to obtain info in that case. +* +* @return pointer to private info +*/ +struct efi_priv *efi_get_priv (void ); + +/** +* efi_set_priv() - Set up a pointer to the EFI-private information +* +* This is called in the stub and app to record the location of this +* information. +* +* @priv: New location of private data +*/ +void efi_set_priv (struct efi_priv *priv); +/** * efi_get_sys_table() - Get access to the main EFI system table *
lib/efi/efi.c +29 -0
@@ -1,5 +1,7 @@// SPDX-License-Identifier: GPL-2.0+ /* +* Functions shared by the app and stub +* * Copyright (c) 2015 Google, Inc * * EFI information obtained here:
@@ -17,6 +19,33 @@#include <efi.h> #include <efi_api.h> +static struct efi_priv *global_priv; + +struct efi_priv *efi_get_priv (void ) +{ +return global_priv; +} + +void efi_set_priv (struct efi_priv *priv) +{ + global_priv = priv; +} + +struct efi_system_table *efi_get_sys_table (void ) +{ +return global_priv->sys_table; +} + +struct efi_boot_services *efi_get_boot (void ) +{ +return global_priv->boot; +} + +unsigned long efi_get_ram_base (void ) +{ +return global_priv->ram_base; +} +/* * Global declaration of gd. *
lib/efi/efi_app.c +2 -19
@@ -27,23 +27,6 @@DECLARE_GLOBAL_DATA_PTR ; -static struct efi_priv *global_priv; - -struct efi_system_table *efi_get_sys_table (void ) -{ -return global_priv->sys_table; -} - -struct efi_boot_services *efi_get_boot (void ) -{ -return global_priv->boot; -} - -unsigned long efi_get_ram_base (void ) -{ -return global_priv->ram_base; -} -int efi_info_get (enum efi_entry_t type,void **datap,int *sizep) {return -ENOSYS ;
@@ -344,7 +327,7 @@ efi_status_t EFIAPI efi_main(efi_handle_t image,/* Set up access to EFI data structures */ efi_init (priv,"App" , image, sys_table); - global_priv = priv; +efi_set_priv (priv);/* * Set up the EFI debug UART so that printf() works. This is
@@ -370,7 +353,7 @@ efi_status_t EFIAPI efi_main(efi_handle_t image,static void efi_exit (void ) { -struct efi_priv *priv = global_priv; +struct efi_priv *priv =efi_get_priv ();free_memory (priv);printf ("U-Boot EFI exiting\n" );
lib/efi/efi_stub.c +4 -3
@@ -31,7 +31,6 @@#error "This file needs to be ported for use on architectures"#endif -static struct efi_priv *global_priv;static bool use_uart;struct __packed desctab_info {
@@ -63,6 +62,8 @@ void _debug_uart_init(void)void putc (const char ch) { +struct efi_priv *priv =efi_get_priv (); +if (ch =='\n' )putc ('\r' );
@@ -73,7 +74,7 @@ void putc(const char ch) ;outb (ch, (ulong)&com_port->thr); }else { -efi_putc (global_priv, ch); +efi_putc (priv, ch); } }
@@ -313,7 +314,7 @@ efi_status_t EFIAPI efi_main(efi_handle_t image,puts (" efi_init() failed\n" );return ret; } - global_priv = priv; +efi_set_priv (priv); cs32 =get_codeseg32 ();if (cs32 <0 )
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 33/39 efi: Move exit_boot_services into a function
In reply to Simon GlassAt present this code is inline in the app and stub. But they do the same thing. The difference is that the stub does it immediately and the app doesn't want to do it until the end (when it boots a kernel) or not at all, if returning to UEFI.
Move it into a function so it can be called as needed.
Also store the memory map so that it can be accessed within the app if needed.
---
Changes in v2:
- Add a sentence about what the patch does
include/efi.h | 32 ++++++++++++++++++++++ lib/efi/efi.c | 68 ++++++++++++++++++++++++++++++++++++++++++++++ lib/efi/efi_app.c | 3 ++ lib/efi/efi_stub.c | 66 ++++++++------------------------------------ 4 files changed, 114 insertions(+), 55 deletions(-)
include/efi.h +32 -0
@@ -407,6 +407,12 @@ static inline struct efi_mem_desc *efi_get_next_mem_desc(* @sys_table: Pointer to system table * @boot: Pointer to boot-services table * @run: Pointer to runtime-services table +* @memmap_key: Key returned from get_memory_map() +* @memmap_desc: List of memory-map records +* @memmap_alloc: Amount of memory allocated for memory map list +* @memmap_size Size of memory-map list in bytes +* @memmap_desc_size: Size of an individual memory-map record, in bytes +* @memmap_version: Memory-map version * * Used by app only: * @use_pool_for_malloc: true if all allocation should go through the EFI 'pool'
@@ -426,6 +432,12 @@ struct efi_priv {struct efi_system_table *sys_table;struct efi_boot_services *boot;struct efi_runtime_services *run; +efi_uintn_t memmap_key; +struct efi_mem_desc *memmap_desc; +efi_uintn_t memmap_alloc; +efi_uintn_t memmap_size; +efi_uintn_t memmap_desc_size; +u32 memmap_version;/* app: */ bool use_pool_for_malloc;
@@ -576,4 +588,24 @@ void efi_putc(struct efi_priv *priv, const char ch);*/ int efi_info_get (enum efi_entry_t type,void **datap,int *sizep); +/** +* efi_store_memory_map() - Collect the memory-map info from EFI +* +* Collect the memory info and store it for later use, e.g. in calling +* exit_boot_services() +* +* @priv: Pointer to private EFI structure +* @return 0 if OK, non-zero on error +*/ +int efi_store_memory_map (struct efi_priv *priv); + +/** +* efi_call_exit_boot_services() - Handlet eh exit-boot-service procedure +* +* Tell EFI we don't want their boot services anymore +* +* @return 0 if OK, non-zero on error +*/ +int efi_call_exit_boot_services (void ); +#endif /* _LINUX_EFI_H */
lib/efi/efi.c +68 -0
71 lines not on this page · raw message
lib/efi/efi_app.c +3 -0
@@ -341,6 +341,9 @@ efi_status_t EFIAPI efi_main(efi_handle_t image,printf ("Failed to set up memory: ret=%lx\n" , ret);return ret; } + ret =efi_store_memory_map (priv); +if (ret) +return ret;printf ("starting\n" );
lib/efi/efi_stub.c +11 -55
98 lines not on this page · raw message
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 34/39 efi: Check for failure when initing the app
In reply to Simon GlassThe stub checks for failure with efi_init(). Add this for the app as well. It is unlikely that anything can be done, but we may as well stop.
---
(no changes since v1)
lib/efi/efi_app.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
lib/efi/efi_app.c +5 -2
@@ -325,8 +325,11 @@ efi_status_t EFIAPI efi_main(efi_handle_t image,efi_status_t ret;/* Set up access to EFI data structures */ -efi_init (priv,"App" , image, sys_table); - + ret =efi_init (priv,"App" , image, sys_table); +if (ret) { +printf ("Failed to set up ARP: err=%lx\n" , ret); +return ret; + }efi_set_priv (priv);/*
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 35/39 efi: Mention that efi_info_get() is only used in the stub
In reply to Simon GlassThis provides access to EFI tables after U-Boot has exited boot services. It is not needed in the app since boot services remain alive and we can just call them whenever needed.
Add a comment to explain this.
---
Changes in v2:
- Fix 'as' typo
include/efi.h | 4 ++++ 1 file changed, 4 insertions(+)
include/efi.h +4 -0
@@ -580,6 +580,10 @@ void efi_putc(struct efi_priv *priv, const char ch);/** * efi_info_get() - get an entry from an EFI table * +* This is called from U-Boot proper to read information set up by the EFI stub. +* It can only be used when running from the EFI stuff, not when U-Boot is +* running as an app. +* * @type: Entry type to search for * @datap: Returns pointer to entry data * @sizep: Returns pointer to entry size
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 36/39 efi: Show when allocated pages are used
In reply to Simon GlassAdd a message here so that both paths of memory allocation are reported.
---
Changes in v2:
- Use log_info() instead of printf()
lib/efi/efi_app.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
lib/efi/efi_app.c +2 -1
@@ -121,13 +121,14 @@ static efi_status_t setup_memory(struct efi_priv *priv) ret = boot->allocate_pages (EFI_ALLOCATE_MAX_ADDRESS , priv->image_data_type, pages, &addr);if (ret) { -printf ("(using pool %lx) " , ret); +log_info ("(using pool %lx) " , ret); priv->ram_base = (ulong )efi_malloc (priv,CONFIG_EFI_RAM_SIZE , &ret);if (!priv->ram_base)return ret; priv->use_pool_for_malloc = true; }else { +log_info ("(using allocated RAM address %lx) " , (ulong )addr); priv->ram_base = addr; } gd->ram_size = pages <<12 ;
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 37/39 efi: Allow easy selection of serial-only operation
In reply to Simon GlassAdd info about how to select vidconsole or serial.
Also set up a demo boot command.
---
Changes in v2:
- Add a better boot command too
include/configs/efi-x86_app.h | 25 +++++++++++++++++++++++++ 1 file changed, 25 insertions(+)
include/configs/efi-x86_app.h +25 -0
@@ -10,8 +10,33 @@#undef CONFIG_TPM_TIS_BASE_ADDRESS +/* +* Select the output device: Put an 'x' prefix before one of these to disable it +*/ + +/* +* Video output - can normally continue after exit_boot_services has been +* called, since output to the display does not require EFI services at that +* point. U-Boot sets up the console memory and does its own drawing. +*/ #define CONFIG_STD_DEVICES_SETTINGS "stdin=serial\0" \ "stdout=vidconsole\0" \ "stderr=vidconsole\0" +/* +* Serial output with no console. Run qemu with: +* +* -display none -serial mon:stdio +* +* This will hang or fail to output on the console after exit_boot_services is +* called. +*/ +#define xCONFIG_STD_DEVICES_SETTINGS "stdin=serial\0" \ + "stdout=serial\0" \ + "stderr=serial\0" + +#undef CONFIG_BOOTCOMMAND + +#define CONFIG_BOOTCOMMAND "part list efi 0; fatls efi 0:1" +#endif
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 38/39 efi: Update efi_get_next_mem_desc() to avoid needing a map
In reply to Simon GlassAt present this function requires a pointer to struct efi_entry_memmap but the only field used in there is the desc_size. We want to be able to use it from the app, so update it to use desc_size directly.
---
(no changes since v1)
arch/x86/cpu/efi/payload.c | 8 ++++---- cmd/efi.c | 34 ++++++++++++++++++---------------- include/efi.h | 4 ++-- 3 files changed, 24 insertions(+), 22 deletions(-)
arch/x86/cpu/efi/payload.c +4 -4
@@ -50,7 +50,7 @@ ulong board_get_usable_ram_top(ulong total_size) end = (struct efi_mem_desc *)((ulong )map + size); desc = map->desc; -for (; desc < end; desc =efi_get_next_mem_desc (map, desc)) { +for (; desc < end; desc =efi_get_next_mem_desc (desc, map->desc_size)) {if (desc->type !=EFI_CONVENTIONAL_MEMORY || desc->physical_start >=1ULL <<32 )continue ;
@@ -88,7 +88,7 @@ int dram_init(void) end = (struct efi_mem_desc *)((ulong )map + size); gd->ram_size =0 ; desc = map->desc; -for (; desc < end; desc =efi_get_next_mem_desc (map, desc)) { +for (; desc < end; desc =efi_get_next_mem_desc (desc, map->desc_size)) {if (desc->type <EFI_MMAP_IO ) gd->ram_size += desc->num_pages <<EFI_PAGE_SHIFT ; }
@@ -113,7 +113,7 @@ int dram_init_banksize(void) desc = map->desc;for (num_banks =0 ; desc < end && num_banks <CONFIG_NR_DRAM_BANKS ; - desc =efi_get_next_mem_desc (map, desc)) { + desc =efi_get_next_mem_desc (desc, map->desc_size)) {/* * We only use conventional memory and ignore * anything less than 1MB.
@@ -196,7 +196,7 @@ unsigned int install_e820_map(unsigned int max_entries, end = (struct efi_mem_desc *)((ulong )map + size);for (desc = map->desc; desc < end; - desc =efi_get_next_mem_desc (map, desc)) { + desc =efi_get_next_mem_desc (desc, map->desc_size)) {if (desc->num_pages ==0 )continue ;
cmd/efi.c +18 -16
@@ -75,16 +75,17 @@ static int h_cmp_entry(const void *v1, const void *v2)/** * efi_build_mem_table() - make a sorted copy of the memory table * -* @map: Pointer to EFI memory map table +* @desc_base: Pointer to EFI memory map table * @size: Size of table in bytes +* @desc_size: Size of each @desc_base record * @skip_bs: True to skip boot-time memory and merge it with conventional * memory. This will significantly reduce the number of table * entries. * Return: pointer to the new table. It should be freed with free() by the * caller. */ -static void *efi_build_mem_table (struct efi_entry_memmap *map,int size, -bool skip_bs) +static void *efi_build_mem_table (struct efi_mem_desc *desc_base,int size, +int desc_size,bool skip_bs) {struct efi_mem_desc *desc, *end, *base, *dest, *prev;int count;
@@ -95,15 +96,16 @@ static void *efi_build_mem_table(struct efi_entry_memmap *map, int size,debug ("%s: Cannot allocate %#x bytes\n" , __func__, size);return NULL ; } - end = (struct efi_mem_desc *)((ulong )map + size); - count = ((ulong )end - (ulong )map->desc) / map->desc_size; -memcpy (base, map->desc, (ulong )end - (ulong )map->desc); -qsort (base, count, map->desc_size, h_cmp_entry); + end = (void *)desc_base + size; + count = ((ulong )end - (ulong )desc_base) / desc_size; +memcpy (base, desc_base, (ulong )end - (ulong )desc_base); +qsort (base, count, desc_size, h_cmp_entry); prev =NULL ; addr =0 ; dest = base; - end = (struct efi_mem_desc *)((ulong )base + count * map->desc_size); -for (desc = base; desc < end; desc =efi_get_next_mem_desc (map, desc)) { + end = (struct efi_mem_desc *)((ulong )base + count * desc_size); +for (desc = base; desc < end; + desc =efi_get_next_mem_desc (desc, desc_size)) {bool merge = true;u32 type = desc->type;
@@ -116,7 +118,7 @@ static void *efi_build_mem_table(struct efi_entry_memmap *map, int size,if (skip_bs &&is_boot_services (desc->type)) type =EFI_CONVENTIONAL_MEMORY ; -memcpy (dest, desc, map->desc_size); +memcpy (dest, desc, desc_size); dest->type = type;if (!skip_bs || !prev) merge = false;
@@ -131,7 +133,7 @@ static void *efi_build_mem_table(struct efi_entry_memmap *map, int size, prev->num_pages += desc->num_pages; }else { prev = dest; - dest =efi_get_next_mem_desc (map, dest); + dest =efi_get_next_mem_desc (dest, desc_size); } addr = desc->physical_start + (desc->num_pages <<EFI_PAGE_SHIFT );
@@ -143,8 +145,8 @@ static void *efi_build_mem_table(struct efi_entry_memmap *map, int size,return base; } -static void efi_print_mem_table (struct efi_entry_memmap *map, -struct efi_mem_desc *desc,bool skip_bs) +static void efi_print_mem_table (struct efi_mem_desc *desc,int desc_size, +bool skip_bs) {u64 attr_seen[ATTR_SEEN_MAX ];int attr_seen_count;
@@ -158,7 +160,7 @@ static void efi_print_mem_table(struct efi_entry_memmap *map, attr_seen_count =0 ; addr =0 ;for (upto =0 ; desc->type !=EFI_MAX_MEMORY_TYPE ; - upto++, desc =efi_get_next_mem_desc (map, desc)) { + upto++, desc =efi_get_next_mem_desc (desc, desc_size)) {const char *name;u64 size;
@@ -238,13 +240,13 @@ static int do_efi_mem(struct cmd_tbl *cmdtp, int flag, int argc, goto done; } - desc =efi_build_mem_table (map, size, skip_bs); + desc =efi_build_mem_table (map->desc, size, map->desc_size, skip_bs);if (!desc) { ret = -ENOMEM ; goto done; } -efi_print_mem_table (map, desc, skip_bs); +efi_print_mem_table (desc, map->desc_size, skip_bs);free (desc); done:if (ret)
include/efi.h +2 -2
@@ -395,9 +395,9 @@ struct efi_entry_systable { };static inline struct efi_mem_desc *efi_get_next_mem_desc ( -struct efi_entry_memmap *map,struct efi_mem_desc *desc) +struct efi_mem_desc *desc,int desc_size) { -return (struct efi_mem_desc *)((ulong )desc + map->desc_size); +return (struct efi_mem_desc *)((ulong )desc + desc_size); }/**
Signature · 1 line
2.33.0.685.g46640cef36-goog
PATCH v2 39/39 efi: Support the efi command in the app
In reply to Simon GlassAt present the 'efi' command only works in the EFI payload. Update it to work in the app too, so the memory map can be examined.
---
Changes in v2:
- Add new patch to support the efi command in the app
cmd/Makefile | 2 +- cmd/efi.c | 46 +++++++++++++++++++++++++++++++--------------- include/efi.h | 15 +++++++++++++++ lib/efi/efi_app.c | 33 +++++++++++++++++++++++++++++++++ 4 files changed, 80 insertions(+), 16 deletions(-)
cmd/Makefile +1 -1
@@ -58,7 +58,7 @@ obj-$(CONFIG_CMD_EXTENSION) += extension_board.o obj-$(CONFIG_CMD_ECHO) += echo.o obj-$(CONFIG_ENV_IS_IN_EEPROM) += eeprom.o obj-$(CONFIG_CMD_EEPROM) += eeprom.o -obj-$(CONFIG_EFI_STUB) += efi.o +obj-$(CONFIG_EFI) += efi.o obj-$(CONFIG_CMD_EFIDEBUG) += efidebug.o obj-$(CONFIG_CMD_ELF) += elf.o obj-$(CONFIG_HUSH_PARSER) += exit.o
cmd/efi.c +31 -15
68 lines not on this page · raw message
include/efi.h +15 -0
@@ -612,4 +612,19 @@ int efi_store_memory_map(struct efi_priv *priv);*/ int efi_call_exit_boot_services (void ); +/** +* efi_get_mmap() - Get the memory map from EFI +* +* This is used in the app. The caller must free *@descp when done +* +* @descp: Returns allocated pointer to EFI memory map table +* @sizep: Returns size of table in bytes +* @keyp: Returns memory-map key +* @desc_sizep: Returns size of each @desc_base record +* @versionp: Returns version number of memory map +* @return 0 on success, -ve on error +*/ +int efi_get_mmap (struct efi_mem_desc **descp,int *sizep,uint *keyp, +int *desc_sizep,uint *versionp); +#endif /* _LINUX_EFI_H */
lib/efi/efi_app.c +33 -0
@@ -32,6 +32,39 @@ int efi_info_get(enum efi_entry_t type, void **datap, int *sizep)return -ENOSYS ; } +int efi_get_mmap (struct efi_mem_desc **descp,int *sizep,uint *keyp, +int *desc_sizep,uint *versionp) +{ +struct efi_priv *priv =efi_get_priv (); +struct efi_boot_services *boot = priv->sys_table->boottime; +efi_uintn_t size, desc_size, key; +struct efi_mem_desc *desc; +efi_status_t ret; +u32 version; + +/* Get the memory map so we can switch off EFI */ + size =0 ; + ret = boot->get_memory_map (&size,NULL , &key, &desc_size, &version); +if (ret !=EFI_BUFFER_TOO_SMALL ) +return log_msg_ret ("get" , -ENOMEM ); + + desc =malloc (size); +if (!desc) +return log_msg_ret ("mem" , -ENOMEM ); + + ret = boot->get_memory_map (&size, desc, &key, &desc_size, &version); +if (ret) +return log_msg_ret ("get" , -EINVAL ); + + *descp = desc; + *sizep = size; + *desc_sizep = desc_size; + *versionp = version; + *keyp = key; + +return 0 ; +} +/** * efi_print_str() - Print a UFT-16 string to the U-Boot console *
Signature · 1 line
2.33.0.685.g46640cef36-goog
Hi Heinrich,
Quoted 176 lines · Simon Glass
At present U-Boot can be built as an EFI app, but it is really just for testing, with very few features. Instead, the payload build is used for booting on top of UEFI, where U-Boot takes over the machine immediately and supplies its own drivers.
But the app could be made more useful.
This series provides access to EFI block devices and the video console within U-Boot. It also restructures the app to make it possible to boot a kernel, although further work is needed to implement this.
This can be thought of as making U-Boot perform a little like 'grub', in that it runs purely based on UEFI-provided services.
Of course a lot more work is required, but this is a start.
Note: It would be very useful to include qemu tests of the app and stub in CI.
This is available at u-boot-dm/efi-working
Changes in v2:
- AddMAINTAINERSentry
- Add a better boot command too
- Add a note thatEFI_GRAPHICS_OUTPUT_PROTOCOLis used
- Add a sentence about what the patch does
- Add a work-around to avoid a toolchain crash
- Add new patch to drop theOF_EMBEDwarning for EFI
- Add new patch to enableDM_ETHfor the app
- Add new patch to support the efi command in the app
- Add new patch with maintainer entry
- Add support for creating a partition table with a filesystem inside
- Add support for running qemu with just a serial console (no display)
- Drop mention of partitions from the commit message
- Expand the commit message to make things clearer
- Fix 'as' typo
- Mention why this patch is included in this series
- Put this patch first so that it is separated from the rest of the series
- Show the correct interface type with 'part list'
- Update commit message to mention that these function are x86-specific
- Update commit message to mention why the data structure is like it is
- Update commit message to mention why ulong is used
- Update documentation
- Update the commit message to explain things better
- Uselog_info()instead ofprintf()Simon Glass (39):
RFC: efi: Drop code that doesn't work with driver model
efi: Add a separate maintainer entry for the app
x86: Keep symbol information in u-boot ELF file
x86: Create a new header for EFI
x86: Show some EFI info with the bdinfo command
x86: Tidy upglobal_datapointer for 64-bit
efi: Add a script for building and testing U-Boot on UEFI
efi: EnableDM_ETHfor the app
efi: Drop theOF_EMBEDwarning for EFI
x86: Create a 32/64-bit selection for the app
efi: Create a 64-bit app
x86: Don't duplicateglobal_ptrin 64-bit EFI app
efi: Add a way to obtain boot services in the app
efi: Add video support to the app
efi: Add EFI uclass for media
efi: Add a media/block driver for EFI block devices
efi: Locate all block devices in the app
patman: Use a ValueError exception iftools.Run()fails
binman: Report an error if test files fail to compile
binman: Support reading the offset of an ELF-file symbol
binman: Allow timeout to occur in the image or its section
binman: Tidy up comments on_DoTestFile()
binman: Support updating the dtb in an ELF file
efi: serial: Support arrow keys
bloblist: Support allocating the bloblist
x86: Allow booting a kernel from the EFI app
x86: Don't process the kernel command line unless enabled
x86: efi: Add room for the binman definition in the dtb
efi: Add comments to structefi_priv
efi: Fixll_boot_init()operation with the app
efi: Add a few comments to the stub
efi: Share structefi_privbetween the app and stub code
efi: Moveexit_boot_servicesinto a function
efi: Check for failure when initing the app
efi: Mention thatefi_info_get()is only used in the stub
efi: Show when allocated pages are used
efi: Allow easy selection of serial-only operation
efi: Updateefi_get_next_mem_desc()to avoid needing a map
efi: Support the efi command in the appMAINTAINERS | 14 + Makefile | 10 +- arch/sandbox/dts/test.dts | 4 + arch/x86/config.mk | 4 +- arch/x86/cpu/efi/payload.c | 21 +- arch/x86/cpu/intel_common/Makefile | 2 +- arch/x86/cpu/u-boot-64.lds | 2 + arch/x86/cpu/x86_64/Makefile | 4 + arch/x86/cpu/x86_64/cpu.c | 32 +- arch/x86/cpu/x86_64/misc.c | 41 +++ arch/x86/dts/Makefile | 2 +- arch/x86/dts/efi-x86_app.dts | 4 + arch/x86/include/asm/efi.h | 39 +++ arch/x86/include/asm/global_data.h | 2 + arch/x86/include/asm/zimage.h | 3 - arch/x86/lib/Makefile | 1 + arch/x86/lib/bdinfo.c | 22 ++ arch/x86/lib/bootm.c | 11 +- arch/x86/lib/zimage.c | 14 +- board/efi/Kconfig | 15 +- board/efi/efi-x86_app/Kconfig | 6 +- board/efi/efi-x86_app/MAINTAINERS | 11 +- cmd/Makefile | 2 +- cmd/efi.c | 76 +++-- common/Kconfig | 15 +- common/bloblist.c | 16 +- common/board_f.c | 8 +- ..._app_defconfig => efi-x86_app32_defconfig} | 3 +- configs/efi-x86_app64_defconfig | 39 +++ disk/part.c | 5 +- doc/develop/bloblist.rst | 16 + doc/develop/uefi/u-boot_on_efi.rst | 74 ++++- drivers/block/Kconfig | 33 +++ drivers/block/Makefile | 4 + drivers/block/blk-uclass.c | 2 +- drivers/block/efi-media-uclass.c | 15 + drivers/block/efi_blk.c | 115 ++++++++ drivers/block/sb_efi_media.c | 20 ++ drivers/serial/serial_efi.c | 11 +- drivers/video/Kconfig | 2 +- drivers/video/efi.c | 45 ++- include/configs/efi-x86_app.h | 31 +- include/dm/uclass-id.h | 1 + include/efi.h | 123 +++++++- include/efi_api.h | 15 + include/init.h | 2 +- lib/efi/Kconfig | 34 ++- lib/efi/efi.c | 106 +++++++ lib/efi/efi_app.c | 276 +++++++++++++++++- lib/efi/efi_stub.c | 95 +++--- lib/efi_driver/Makefile | 2 +- lib/efi_loader/Kconfig | 1 + lib/efi_loader/efi_device_path.c | 96 ++---- lib/efi_loader/efi_disk.c | 48 --- scripts/build-efi.sh | 188 ++++++++++++ test/dm/Makefile | 1 + test/dm/efi_media.c | 24 ++ tools/binman/binman.rst | 36 +++ tools/binman/cmdline.py | 2 + tools/binman/control.py | 11 + tools/binman/elf.py | 74 ++++- tools/binman/elf_test.py | 45 ++- tools/binman/ftest.py | 91 +++++- tools/binman/test/Makefile | 13 +- tools/binman/test/bss_data.c | 2 +- tools/binman/test/embed_data.c | 16 + tools/binman/test/embed_data.lds | 23 ++ tools/binman/test/u_boot_binman_embed.c | 13 + tools/binman/test/u_boot_binman_embed.lds | 29 ++ tools/binman/test/u_boot_binman_embed_sm.c | 13 + tools/patman/tools.py | 2 +- 71 files changed, 1841 insertions(+), 337 deletions(-) create mode 100644 arch/x86/cpu/x86_64/misc.c create mode 100644 arch/x86/include/asm/efi.h create mode 100644 arch/x86/lib/bdinfo.c rename configs/{efi-x86_app_defconfig => efi-x86_app32_defconfig} (94%) create mode 100644 configs/efi-x86_app64_defconfig create mode 100644 drivers/block/efi-media-uclass.c create mode 100644 drivers/block/efi_blk.c create mode 100644 drivers/block/sb_efi_media.c create mode 100755 scripts/build-efi.sh create mode 100644 test/dm/efi_media.c create mode 100644 tools/binman/test/embed_data.c create mode 100644 tools/binman/test/embed_data.lds create mode 100644 tools/binman/test/u_boot_binman_embed.c create mode 100644 tools/binman/test/u_boot_binman_embed.lds create mode 100644 tools/binman/test/u_boot_binman_embed_sm.c
Quoted 3 lines
--
2.33.0.685.g46640cef36-goog
These patches seem to be in your queue. Can you please give me an update?
Thanks,
Simon