event: Provide support for events to connect subsystems
Subsystems (11)
v2 cover event: Provide support for events to connect subsystems
It is a common need in U-Boot to have one subsystem notify another when something happens. An example is reading a partition table when a new block device is set up.
It is also common to add weak functions and 'hook' functions to modify how U-Boot works. See for example ft_board_setup() and the like.
U-Boot would benefit from a generic mechanism to handle these cases, with the ability to hook into various 'events' in a subsystem-independent and transparent way.
This series provides a way to create and dispatch events, with a way of registering a 'spy' which watches for events of different types. This allows 'hook' functions to be created in a generic way.
It also includes a script to list the hooks in an image, which is a bit easier to debug than weak functions, as well as an 'event' command to do the same from within U-Boot.
These 'static' events can be used to replace hooks like misc_init_f(), for example. Also included is basic support for 'dynamic' events, where a spy can be registered at runtime. The need for this is still being figured out.
Changes in v2:
- Add a 'used' attribute to avoid LTO dropping the code
- Update keymile pg-wcom boards also
- Make the sandbox function static
- Convert arch/arm/mach-imx/imx8/cpu.c also
- Add DM_EVENT to ARCH_IMX8 too
- Tidy up galileo defconfig and rename quark_init_dm()
- Make a few more of the functions static
- Add a weak function for spl
- Update for patman snake-case change
- Use a regular expression in the test to avoid dependency on build dir
Simon Glass (12):
Makefile: Allow LTO to be disabled for a build
sandbox: start: Sort the header files
binman: Expand elf support a little
event: Add basic support for events
event: Add a simple test
event: Set up the event system on start-up
event: Add events for device probe/remove
event: Convert misc_init_f() to use events
event: Convert arch_cpu_init_dm() to use events
event: Add a command
event: Add a script to decode the event-spy list
event: Add documentation
Tim Harvey (1):
phy: nop-phy: Fix phy reset if no reset-gpio defined
MAINTAINERS | 10 + Makefile | 18 +- arch/Kconfig | 3 + arch/arm/Kconfig | 4 + arch/arm/config.mk | 4 +- arch/arm/include/asm/global_data.h | 2 +- arch/arm/mach-imx/imx8/cpu.c | 4 +- arch/arm/mach-imx/imx8m/soc.c | 4 +- arch/arm/mach-imx/imx8ulp/soc.c | 4 +- arch/arm/mach-omap2/am33xx/board.c | 4 +- arch/arm/mach-omap2/hwinit-common.c | 5 +- arch/mips/Kconfig | 1 + arch/mips/mach-pic32/cpu.c | 4 +- arch/nios2/cpu/cpu.c | 4 +- arch/riscv/cpu/cpu.c | 5 +- arch/riscv/include/asm/system.h | 5 + arch/riscv/lib/spl.c | 3 +- arch/sandbox/cpu/start.c | 8 +- arch/x86/cpu/baytrail/cpu.c | 4 +- arch/x86/cpu/broadwell/cpu.c | 4 +- arch/x86/cpu/ivybridge/cpu.c | 4 +- arch/x86/cpu/quark/quark.c | 4 +- arch/x86/include/asm/fsp2/fsp_api.h | 8 + arch/x86/lib/fsp2/fsp_init.c | 4 +- arch/x86/lib/spl.c | 5 +- arch/x86/lib/tpl.c | 10 - board/google/chromebook_coral/coral.c | 7 +- board/keymile/kmcent2/kmcent2.c | 4 +- .../keymile/pg-wcom-ls102xa/pg-wcom-ls102xa.c | 5 +- cmd/Kconfig | 8 + cmd/Makefile | 1 + cmd/event.c | 27 +++ common/Kconfig | 37 ++- common/Makefile | 2 + common/board_f.c | 13 +- common/board_r.c | 1 + common/event.c | 196 ++++++++++++++++ common/log.c | 1 + configs/chromebook_coral_defconfig | 1 + configs/kmcent2_defconfig | 2 +- configs/pg_wcom_expu1_defconfig | 1 + configs/pg_wcom_expu1_update_defconfig | 1 + configs/pg_wcom_seli8_defconfig | 1 + configs/pg_wcom_seli8_update_defconfig | 1 + configs/sandbox64_defconfig | 1 - configs/sandbox_defconfig | 1 - configs/sandbox_flattree_defconfig | 1 - configs/sandbox_spl_defconfig | 1 - configs/tools-only_defconfig | 1 - doc/build/gcc.rst | 17 ++ doc/develop/event.rst | 66 ++++++ doc/develop/index.rst | 1 + doc/usage/event.rst | 49 ++++ doc/usage/index.rst | 1 + drivers/core/Kconfig | 10 + drivers/core/device-remove.c | 8 + drivers/core/device.c | 9 + drivers/core/root.c | 5 + drivers/phy/nop-phy.c | 12 +- include/asm-generic/global_data.h | 13 ++ include/configs/km/pg-wcom-ls102xa.h | 2 - include/dm/device-internal.h | 10 + include/event.h | 210 ++++++++++++++++++ include/event_internal.h | 35 +++ include/init.h | 12 - include/log.h | 2 + scripts/event_dump.py | 115 ++++++++++ test/common/Makefile | 1 + test/common/event.c | 85 +++++++ test/py/tests/test_event_dump.py | 20 ++ test/test-main.c | 4 + tools/binman/elf.py | 58 ++++- 72 files changed, 1108 insertions(+), 86 deletions(-) create mode 100644 cmd/event.c create mode 100644 common/event.c create mode 100644 doc/develop/event.rst create mode 100644 doc/usage/event.rst create mode 100644 include/event.h create mode 100644 include/event_internal.h create mode 100755 scripts/event_dump.py create mode 100644 test/common/event.c create mode 100644 test/py/tests/test_event_dump.py
Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
PATCH v2 1/13 phy: nop-phy: Fix phy reset if no reset-gpio defined
From: Tim Harvey <[email protected]>
Ensure there is a valid reset-gpio defined before using it.
---
(no changes since v1)
drivers/phy/nop-phy.c | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-)
drivers/phy/nop-phy.c +7 -5
@@ -45,11 +45,13 @@ static int nop_phy_init(struct phy *phy)#if CONFIG_IS_ENABLED (DM_GPIO )/* Take phy out of reset */ - ret =dm_gpio_set_value (&priv->reset_gpio, false); -if (ret) { -if (CONFIG_IS_ENABLED (CLK )) -clk_disable_bulk (&priv->bulk); -return ret; +if (dm_gpio_is_valid (&priv->reset_gpio)) { + ret =dm_gpio_set_value (&priv->reset_gpio, false); +if (ret) { +if (CONFIG_IS_ENABLED (CLK )) +clk_disable_bulk (&priv->bulk); +return ret; + } }#endif return 0 ;
Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
Re: [PATCH v2 01/13] phy: nop-phy: Fix phy reset if no reset-gpio defined
Quoted 8 lines · Simon Glass
From: Tim Harvey<[email protected]>
Ensure there is a valid reset-gpio defined before using it.
Fixes f9852acdce02 ("phy: nop-phy: Fix enabling reset")Cc Adam Ford<[email protected]>Signed-off-by Tim Harvey<[email protected]>Signed-off-by Simon Glass<[email protected]>
This patch is already in origin/master.
PATCH v2 2/13 Makefile: Allow LTO to be disabled for a build
In reply to Simon GlassLTO (Link-Time Optimisation) is an very useful feature which can significantly reduce the size of U-Boot binaries. So far it has been made available for selected ARM boards and sandbox.
However, incremental builds are much slower when LTO is used. For example, an incremental build of sandbox takes 2.1 seconds on my machine, but 6.7 seconds with LTO enabled.
Add a LTO_BUILD=n parameter to the build, so it can be disabled during development if needed, for faster builds.
Add some documentation about LTO while we are here.
---
(no changes since v1)
Makefile | 18 +++++++++++++----- arch/arm/config.mk | 4 ++-- arch/arm/include/asm/global_data.h | 2 +- doc/build/gcc.rst | 17 +++++++++++++++++ 4 files changed, 33 insertions(+), 8 deletions(-)
Makefile +13 -5
@@ -434,6 +434,9 @@ KBUILD_CFLAGS += -fshort-wchar -fno-strict-aliasing KBUILD_AFLAGS := -D__ASSEMBLY__ KBUILD_LDFLAGS := +# Set this to "n" use of LTO for this build, e.g. LTO_BUILD=n +LTO_BUILD ?= y + ifeq ($(cc-name),clang) ifneq ($(CROSS_COMPILE),) CLANG_TARGET := --target=$(notdir $(CROSS_COMPILE:%-=%))
@@ -643,6 +646,11 @@ export CFLAGS_EFI # Compiler flags to add when building EFI app export CFLAGS_NON_EFI # Compiler flags to remove when building EFI app export EFI_TARGET # binutils target if EFI is natively supported +export LTO_ENABLE + +# This is y if LTO is enabled for this build +LTO_ENABLE=$(if $(CONFIG_LTO),$(LTO_BUILD),) + # If board code explicitly specified LDSCRIPT or CONFIG_SYS_LDSCRIPT, use # that (or fail if absent). Otherwise, search for a linker script in a # standard location.
@@ -690,16 +698,16 @@ endif LTO_CFLAGS := LTO_FINAL_LDFLAGS := export LTO_CFLAGS LTO_FINAL_LDFLAGS -ifdef CONFIG_LTO +ifeq ($(LTO_ENABLE),y) ifeq ($(cc-name),clang) - LTO_CFLAGS += -flto + LTO_CFLAGS += -DLTO_ENABLE -flto LTO_FINAL_LDFLAGS += -flto AR = $(shell $(CC) -print-prog-name=llvm-ar) NM = $(shell $(CC) -print-prog-name=llvm-nm) else NPROC := $(shell nproc 2>/dev/null || echo 1) - LTO_CFLAGS += -flto=$(NPROC) + LTO_CFLAGS += -DLTO_ENABLE -flto=$(NPROC) LTO_FINAL_LDFLAGS += -fuse-linker-plugin -flto=$(NPROC) # use plugin aware tools
@@ -1740,7 +1748,7 @@ ARCH_POSTLINK := $(wildcard $(srctree)/arch/$(ARCH)/Makefile.postlink) # Generate linker list symbols references to force compiler to not optimize # them away when compiling with LTO -ifdef CONFIG_LTO +ifeq ($(LTO_ENABLE),y) u-boot-keep-syms-lto := keep-syms-lto.o u-boot-keep-syms-lto_c := $(patsubst %.o,%.c,$(u-boot-keep-syms-lto))
@@ -1762,7 +1770,7 @@ endif # Rule to link u-boot # May be overridden by arch/$(ARCH)/config.mk -ifdef CONFIG_LTO +ifeq ($(LTO_ENABLE),y) quiet_cmd_u-boot__ ?= LTO $@ cmd_u-boot__ ?= \ $(CC) -nostdlib -nostartfiles \
arch/arm/config.mk +2 -2
@@ -15,11 +15,11 @@ CFLAGS_NON_EFI := -fno-pic -ffixed-r9 -ffunction-sections -fdata-sections \ -fstack-protector-strong CFLAGS_EFI := -fpic -fshort-wchar -ifneq ($(CONFIG_LTO)$(CONFIG_USE_PRIVATE_LIBGCC),yy) +ifneq ($(LTO_ENABLE)$(CONFIG_USE_PRIVATE_LIBGCC),yy) LDFLAGS_FINAL += --gc-sections endif -ifndef CONFIG_LTO +ifneq ($(LTO_ENABLE),y) PLATFORM_RELFLAGS += -ffunction-sections -fdata-sections endif
arch/arm/include/asm/global_data.h +1 -1
@@ -98,7 +98,7 @@ struct arch_global_data {#include <asm-generic/global_data.h> -#if defined(__clang__) || defined(CONFIG_LTO ) +#if defined(__clang__) || defined(LTO_ENABLE )#define DECLARE_GLOBAL_DATA_PTR #define gd get_gd()
doc/build/gcc.rst +17 -0
@@ -151,6 +151,23 @@ of dtc is new enough. It also makes sure that pylibfdt is present, if needed Note that the :doc:`tools` are always built with the included version of libfdt so it is not possible to build U-Boot tools with a system libfdt, at present. +Link-time optimisation (LTO) +~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +U-Boot supports link-time optimisation which can reduce the size of the final +U-Boot binaries, particularly with SPL. + +At present this can be enabled by ARM boards by adding `CONFIG_LTO=y` into the +defconfig file. Other architectures are not supported. LTO is enabled by default +for sandbox. + +This does incur a link-time penalty of several seconds. For faster incremental +builds during development, you can disable it by setting `LTO_BUILD` to `n`. + +.. code-block:: bash + + LTO_BUILD=n make + Other build targets ~~~~~~~~~~~~~~~~~~~
Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
Re: [PATCH v2 02/13] Makefile: Allow LTO to be disabled for a build
Quoted 8 lines · Simon Glass
LTO (Link-Time Optimisation) is an very useful feature which can significantly reduce the size of U-Boot binaries. So far it has been made available for selected ARM boards and sandbox.
However, incremental builds are much slower when LTO is used. For example, an incremental build of sandbox takes 2.1 seconds on my machine, but 6.7 seconds with LTO enabled.
Quoted 6 lines
Add a LTO_BUILD=n parameter to the build, so it can be disabled during development if needed, for faster builds.
Add some documentation about LTO while we are here.
Signed-off-by Simon Glass <[email protected]>
We don't need this since you can do:
make EXTRA_CFLAGS="-fno-lto" EXTRA_LDFLAGS="-fno-lto"
to pass -fno-lto to compile/linking and disable lto and per
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46905 this has been working for some time.
Not that you need to respin the series for this.
Signature · 1 line
Tom
Re: [PATCH v2 02/13] Makefile: Allow LTO to be disabled for a build
Hi Tom,
Quoted 16 lines · Tom Rini
Quoted 13 lines · Simon Glass
LTO (Link-Time Optimisation) is an very useful feature which can significantly reduce the size of U-Boot binaries. So far it has been made available for selected ARM boards and sandbox.
However, incremental builds are much slower when LTO is used. For example, an incremental build of sandbox takes 2.1 seconds on my machine, but 6.7 seconds with LTO enabled.
Add a LTO_BUILD=n parameter to the build, so it can be disabled during development if needed, for faster builds.
Add some documentation about LTO while we are here.
Quoted 7 lines
Quoted 1 line
Signed-off-by Simon Glass <[email protected]>We don't need this since you can do:
make EXTRA_CFLAGS="-fno-lto" EXTRA_LDFLAGS="-fno-lto"
to pass -fno-lto to compile/linking and disable lto and per
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46905 this has been working for some time.
Thanks for that, it is a big pain point for me, picking up this patch for every series I write. The incremental build time for sandbox goes from 3 seconds to 27 seconds on my laptop with LTO, which is intolerable.
The EXTRA_CFLAGS says it is for 'Backward compatibility' and it still does the various LTO things (i.e. it changes the build logic). It seems odd to me to enable the option and then disable it later in the command line. It is therefore not quite equivalent. But it seems to work well enough for me fom a small amount of testing. If you are really set on not having a special option for it, I can live with it for now. I'm also not convinced that my patch entirely removes the LTO stuff in a consistent way.
Quoted 1 line
Not that you need to respin the series for this.
Regards,
Simon
Re: [PATCH v2 02/13] Makefile: Allow LTO to be disabled for a build
Quoted 27 lines · Simon Glass
Hi Tom,
Quoted 16 lines · Tom Rini
Quoted 13 lines · Simon Glass
LTO (Link-Time Optimisation) is an very useful feature which can significantly reduce the size of U-Boot binaries. So far it has been made available for selected ARM boards and sandbox.
However, incremental builds are much slower when LTO is used. For example, an incremental build of sandbox takes 2.1 seconds on my machine, but 6.7 seconds with LTO enabled.
Add a LTO_BUILD=n parameter to the build, so it can be disabled during development if needed, for faster builds.
Add some documentation about LTO while we are here.
Quoted 7 lines
Quoted 1 line
Signed-off-by Simon Glass <[email protected]>We don't need this since you can do:
make EXTRA_CFLAGS="-fno-lto" EXTRA_LDFLAGS="-fno-lto"
to pass -fno-lto to compile/linking and disable lto and per
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46905 this has been working for some time.
Quoted 4 lines
Thanks for that, it is a big pain point for me, picking up this patch for every series I write. The incremental build time for sandbox goes from 3 seconds to 27 seconds on my laptop with LTO, which is intolerable.
Yeah, I noticed it's visible on my laptop, but not at all on my desktop (i7 vs Ryzen 7).
Quoted 2 lines
The
EXTRA_CFLAGSsays it is for 'Backward compatibility' and it still does the various LTO things (i.e. it changes the build logic). It
We're unlikely to move to newer Linux kernel kbuild logic so this isn't going away, and there's not much in the way of logic that's changed for LTO that I see.
Quoted 6 lines
seems odd to me to enable the option and then disable it later in the command line. It is therefore not quite equivalent. But it seems to work well enough for me fom a small amount of testing. If you are really set on not having a special option for it, I can live with it for now. I'm also not convinced that my patch entirely removes the LTO stuff in a consistent way.
Yeah, I really don't want to go down the path of overriding CONFIG options via make/environment logic. I'm also open to turning off LTO on sandbox and on with qemu-* so it gets wider CI testing.
Signature · 1 line
Tom
Re: [PATCH v2 02/13] Makefile: Allow LTO to be disabled for a build
Hi Tom,
Quoted 51 lines · Tom Rini
Quoted 27 lines · Simon Glass
Hi Tom,
Quoted 16 lines · Tom Rini
Quoted 13 lines · Simon Glass
LTO (Link-Time Optimisation) is an very useful feature which can significantly reduce the size of U-Boot binaries. So far it has been made available for selected ARM boards and sandbox.
However, incremental builds are much slower when LTO is used. For example, an incremental build of sandbox takes 2.1 seconds on my machine, but 6.7 seconds with LTO enabled.
Add a LTO_BUILD=n parameter to the build, so it can be disabled during development if needed, for faster builds.
Add some documentation about LTO while we are here.
Quoted 7 lines
Quoted 1 line
Signed-off-by Simon Glass <[email protected]>We don't need this since you can do:
make EXTRA_CFLAGS="-fno-lto" EXTRA_LDFLAGS="-fno-lto"
to pass -fno-lto to compile/linking and disable lto and per
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46905 this has been working for some time.Quoted 4 lines
Thanks for that, it is a big pain point for me, picking up this patch for every series I write. The incremental build time for sandbox goes from 3 seconds to 27 seconds on my laptop with LTO, which is intolerable.
Yeah, I noticed it's visible on my laptop, but not at all on my desktop (i7 vs Ryzen 7).
Quoted 2 lines
The
EXTRA_CFLAGSsays it is for 'Backward compatibility' and it still does the various LTO things (i.e. it changes the build logic). ItWe're unlikely to move to newer Linux kernel kbuild logic so this isn't going away, and there's not much in the way of logic that's changed for LTO that I see.
Quoted 6 lines
seems odd to me to enable the option and then disable it later in the command line. It is therefore not quite equivalent. But it seems to work well enough for me fom a small amount of testing. If you are really set on not having a special option for it, I can live with it for now. I'm also not convinced that my patch entirely removes the LTO stuff in a consistent way.
Quoted 3 lines
Yeah, I really don't want to go down the path of overriding CONFIG options via make/environment logic. I'm also open to turning off LTO on sandbox and on with qemu-* so it gets wider CI testing.
Yes you did mention that, but the problem is that LTO is very handy with sandbox, to test the strange things that happen. For example, I found the bug where LTO was dropping a linker-list item, using sandbox. We could perhaps make one of the sandbox builds not use LTO, e.g. sandbox_flattree ?
Regards,
Simon
Re: [PATCH v2 02/13] Makefile: Allow LTO to be disabled for a build
Quoted 58 lines · Simon Glass
Hi Tom,
Quoted 51 lines · Tom Rini
Quoted 27 lines · Simon Glass
Hi Tom,
Quoted 16 lines · Tom Rini
Quoted 13 lines · Simon Glass
LTO (Link-Time Optimisation) is an very useful feature which can significantly reduce the size of U-Boot binaries. So far it has been made available for selected ARM boards and sandbox.
However, incremental builds are much slower when LTO is used. For example, an incremental build of sandbox takes 2.1 seconds on my machine, but 6.7 seconds with LTO enabled.
Add a LTO_BUILD=n parameter to the build, so it can be disabled during development if needed, for faster builds.
Add some documentation about LTO while we are here.
Quoted 7 lines
Quoted 1 line
Signed-off-by Simon Glass <[email protected]>We don't need this since you can do:
make EXTRA_CFLAGS="-fno-lto" EXTRA_LDFLAGS="-fno-lto"
to pass -fno-lto to compile/linking and disable lto and per
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46905 this has been working for some time.Quoted 4 lines
Thanks for that, it is a big pain point for me, picking up this patch for every series I write. The incremental build time for sandbox goes from 3 seconds to 27 seconds on my laptop with LTO, which is intolerable.
Yeah, I noticed it's visible on my laptop, but not at all on my desktop (i7 vs Ryzen 7).
Quoted 2 lines
The
EXTRA_CFLAGSsays it is for 'Backward compatibility' and it still does the various LTO things (i.e. it changes the build logic). ItWe're unlikely to move to newer Linux kernel kbuild logic so this isn't going away, and there's not much in the way of logic that's changed for LTO that I see.
Quoted 6 lines
seems odd to me to enable the option and then disable it later in the command line. It is therefore not quite equivalent. But it seems to work well enough for me fom a small amount of testing. If you are really set on not having a special option for it, I can live with it for now. I'm also not convinced that my patch entirely removes the LTO stuff in a consistent way.
Quoted 3 lines
Yeah, I really don't want to go down the path of overriding CONFIG options via make/environment logic. I'm also open to turning off LTO on sandbox and on with qemu-* so it gets wider CI testing.
Quoted 5 lines
Yes you did mention that, but the problem is that LTO is very handy with sandbox, to test the strange things that happen. For example, I found the bug where LTO was dropping a linker-list item, using sandbox. We could perhaps make one of the sandbox builds not use LTO, e.g.
sandbox_flattree?
Well, the big issue with LTO+sandbox is that it slows down your workflow, so I would think you want the inverse, one platform does enable it?
Signature · 1 line
Tom
Re: [PATCH v2 02/13] Makefile: Allow LTO to be disabled for a build
Hi Tom,
Quoted 66 lines · Tom Rini
Quoted 58 lines · Simon Glass
Hi Tom,
Quoted 51 lines · Tom Rini
Quoted 27 lines · Simon Glass
Hi Tom,
Quoted 16 lines · Tom Rini
Quoted 13 lines · Simon Glass
LTO (Link-Time Optimisation) is an very useful feature which can significantly reduce the size of U-Boot binaries. So far it has been made available for selected ARM boards and sandbox.
However, incremental builds are much slower when LTO is used. For example, an incremental build of sandbox takes 2.1 seconds on my machine, but 6.7 seconds with LTO enabled.
Add a LTO_BUILD=n parameter to the build, so it can be disabled during development if needed, for faster builds.
Add some documentation about LTO while we are here.
Quoted 7 lines
Quoted 1 line
Signed-off-by Simon Glass <[email protected]>We don't need this since you can do:
make EXTRA_CFLAGS="-fno-lto" EXTRA_LDFLAGS="-fno-lto"
to pass -fno-lto to compile/linking and disable lto and per
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46905 this has been working for some time.Quoted 4 lines
Thanks for that, it is a big pain point for me, picking up this patch for every series I write. The incremental build time for sandbox goes from 3 seconds to 27 seconds on my laptop with LTO, which is intolerable.
Yeah, I noticed it's visible on my laptop, but not at all on my desktop (i7 vs Ryzen 7).
Quoted 2 lines
The
EXTRA_CFLAGSsays it is for 'Backward compatibility' and it still does the various LTO things (i.e. it changes the build logic). ItWe're unlikely to move to newer Linux kernel kbuild logic so this isn't going away, and there's not much in the way of logic that's changed for LTO that I see.
Quoted 6 lines
seems odd to me to enable the option and then disable it later in the command line. It is therefore not quite equivalent. But it seems to work well enough for me fom a small amount of testing. If you are really set on not having a special option for it, I can live with it for now. I'm also not convinced that my patch entirely removes the LTO stuff in a consistent way.
Quoted 3 lines
Yeah, I really don't want to go down the path of overriding CONFIG options via make/environment logic. I'm also open to turning off LTO on sandbox and on with qemu-* so it gets wider CI testing.
Quoted 5 lines
Yes you did mention that, but the problem is that LTO is very handy with sandbox, to test the strange things that happen. For example, I found the bug where LTO was dropping a linker-list item, using sandbox. We could perhaps make one of the sandbox builds not use LTO, e.g.
sandbox_flattree?
Quoted 3 lines
Well, the big issue with LTO+sandbox is that it slows down your workflow, so I would think you want the inverse, one platform does enable it?
It slows down all boards that use it, actually, so I normally don't want it enabled for my IDE. This is not sandbox-specific.
Regards,
Simon
Re: [PATCH v2 02/13] Makefile: Allow LTO to be disabled for a build
Quoted 69 lines · Simon Glass
Hi Tom,
Quoted 66 lines · Tom Rini
Quoted 63 lines · Simon Glass
Hi Tom,
Quoted 51 lines · Tom Rini
Quoted 27 lines · Simon Glass
Hi Tom,
Quoted 16 lines · Tom Rini
Quoted 13 lines · Simon Glass
LTO (Link-Time Optimisation) is an very useful feature which can significantly reduce the size of U-Boot binaries. So far it has been made available for selected ARM boards and sandbox.
However, incremental builds are much slower when LTO is used. For example, an incremental build of sandbox takes 2.1 seconds on my machine, but 6.7 seconds with LTO enabled.
Add a LTO_BUILD=n parameter to the build, so it can be disabled during development if needed, for faster builds.
Add some documentation about LTO while we are here.
Quoted 7 lines
Quoted 1 line
Signed-off-by Simon Glass <[email protected]>We don't need this since you can do:
make EXTRA_CFLAGS="-fno-lto" EXTRA_LDFLAGS="-fno-lto"
to pass -fno-lto to compile/linking and disable lto and per
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46905 this has been working for some time.Quoted 4 lines
Thanks for that, it is a big pain point for me, picking up this patch for every series I write. The incremental build time for sandbox goes from 3 seconds to 27 seconds on my laptop with LTO, which is intolerable.
Yeah, I noticed it's visible on my laptop, but not at all on my desktop (i7 vs Ryzen 7).
Quoted 2 lines
The
EXTRA_CFLAGSsays it is for 'Backward compatibility' and it still does the various LTO things (i.e. it changes the build logic). ItWe're unlikely to move to newer Linux kernel kbuild logic so this isn't going away, and there's not much in the way of logic that's changed for LTO that I see.
Quoted 6 lines
seems odd to me to enable the option and then disable it later in the command line. It is therefore not quite equivalent. But it seems to work well enough for me fom a small amount of testing. If you are really set on not having a special option for it, I can live with it for now. I'm also not convinced that my patch entirely removes the LTO stuff in a consistent way.
Quoted 3 lines
Yeah, I really don't want to go down the path of overriding CONFIG options via make/environment logic. I'm also open to turning off LTO on sandbox and on with qemu-* so it gets wider CI testing.
Yes you did mention that, but the problem is that LTO is very handy with sandbox, to test the strange things that happen. For example, I found the bug where LTO was dropping a linker-list item, using sandbox. We could perhaps make one of the sandbox builds not use LTO, e.g.
sandbox_flattree?
Quoted 6 lines
Quoted 3 lines
Well, the big issue with LTO+sandbox is that it slows down your workflow, so I would think you want the inverse, one platform does enable it?
It slows down all boards that use it, actually, so I normally don't want it enabled for my IDE. This is not sandbox-specific.
The degree of slowdown depends on what you're building on. I see several seconds of link time on my laptop and nothing on my desktop (which isn't that high end either, it's not an EPYC or anything) when LTO is/isn't enabled.
Signature · 1 line
Tom
Re: [PATCH v2 02/13] Makefile: Allow LTO to be disabled for a build
Hi Tom,
Quoted 75 lines · Tom Rini
Quoted 73 lines · Simon Glass
Hi Tom,
Quoted 69 lines · Tom Rini
Quoted 58 lines · Simon Glass
Hi Tom,
Quoted 51 lines · Tom Rini
Quoted 27 lines · Simon Glass
Hi Tom,
Quoted 16 lines · Tom Rini
Quoted 13 lines · Simon Glass
LTO (Link-Time Optimisation) is an very useful feature which can significantly reduce the size of U-Boot binaries. So far it has been made available for selected ARM boards and sandbox.
However, incremental builds are much slower when LTO is used. For example, an incremental build of sandbox takes 2.1 seconds on my machine, but 6.7 seconds with LTO enabled.
Add a LTO_BUILD=n parameter to the build, so it can be disabled during development if needed, for faster builds.
Add some documentation about LTO while we are here.
Quoted 7 lines
Quoted 1 line
Signed-off-by Simon Glass <[email protected]>We don't need this since you can do:
make EXTRA_CFLAGS="-fno-lto" EXTRA_LDFLAGS="-fno-lto"
to pass -fno-lto to compile/linking and disable lto and per
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46905 this has been working for some time.Quoted 4 lines
Thanks for that, it is a big pain point for me, picking up this patch for every series I write. The incremental build time for sandbox goes from 3 seconds to 27 seconds on my laptop with LTO, which is intolerable.
Yeah, I noticed it's visible on my laptop, but not at all on my desktop (i7 vs Ryzen 7).
Quoted 2 lines
The
EXTRA_CFLAGSsays it is for 'Backward compatibility' and it still does the various LTO things (i.e. it changes the build logic). ItWe're unlikely to move to newer Linux kernel kbuild logic so this isn't going away, and there's not much in the way of logic that's changed for LTO that I see.
Quoted 6 lines
seems odd to me to enable the option and then disable it later in the command line. It is therefore not quite equivalent. But it seems to work well enough for me fom a small amount of testing. If you are really set on not having a special option for it, I can live with it for now. I'm also not convinced that my patch entirely removes the LTO stuff in a consistent way.
Quoted 3 lines
Yeah, I really don't want to go down the path of overriding CONFIG options via make/environment logic. I'm also open to turning off LTO on sandbox and on with qemu-* so it gets wider CI testing.
Quoted 5 lines
Yes you did mention that, but the problem is that LTO is very handy with sandbox, to test the strange things that happen. For example, I found the bug where LTO was dropping a linker-list item, using sandbox. We could perhaps make one of the sandbox builds not use LTO, e.g.
sandbox_flattree?Well, the big issue with LTO+sandbox is that it slows down your workflow, so I would think you want the inverse, one platform does enable it?
Quoted 7 lines
Quoted 2 lines
It slows down all boards that use it, actually, so I normally don't want it enabled for my IDE. This is not sandbox-specific.
The degree of slowdown depends on what you're building on. I see several seconds of link time on my laptop and nothing on my desktop (which isn't that high end either, it's not an EPYC or anything) when LTO is/isn't enabled.
For me, on 16-core AMD desktop, incremental build
LTO: real 0m7.744s
No LTO: real 0m3.172s
On 4-core laptop:
No LTO: real 0m2.429s
LTO: real 0m15.650s
Perhaps the disconnect here is that you just don't see any difference? What times are you seeing?
Regards,
SImon
Re: [PATCH v2 02/13] Makefile: Allow LTO to be disabled for a build
Quoted 91 lines · Simon Glass
Hi Tom,
Quoted 75 lines · Tom Rini
Quoted 73 lines · Simon Glass
Hi Tom,
Quoted 69 lines · Tom Rini
Quoted 58 lines · Simon Glass
Hi Tom,
Quoted 51 lines · Tom Rini
Quoted 27 lines · Simon Glass
Hi Tom,
Quoted 16 lines · Tom Rini
Quoted 13 lines · Simon Glass
LTO (Link-Time Optimisation) is an very useful feature which can significantly reduce the size of U-Boot binaries. So far it has been made available for selected ARM boards and sandbox.
However, incremental builds are much slower when LTO is used. For example, an incremental build of sandbox takes 2.1 seconds on my machine, but 6.7 seconds with LTO enabled.
Add a LTO_BUILD=n parameter to the build, so it can be disabled during development if needed, for faster builds.
Add some documentation about LTO while we are here.
Quoted 7 lines
Quoted 1 line
Signed-off-by Simon Glass <[email protected]>We don't need this since you can do:
make EXTRA_CFLAGS="-fno-lto" EXTRA_LDFLAGS="-fno-lto"
to pass -fno-lto to compile/linking and disable lto and per
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46905 this has been working for some time.Quoted 4 lines
Thanks for that, it is a big pain point for me, picking up this patch for every series I write. The incremental build time for sandbox goes from 3 seconds to 27 seconds on my laptop with LTO, which is intolerable.
Yeah, I noticed it's visible on my laptop, but not at all on my desktop (i7 vs Ryzen 7).
Quoted 2 lines
The
EXTRA_CFLAGSsays it is for 'Backward compatibility' and it still does the various LTO things (i.e. it changes the build logic). ItWe're unlikely to move to newer Linux kernel kbuild logic so this isn't going away, and there's not much in the way of logic that's changed for LTO that I see.
Quoted 6 lines
seems odd to me to enable the option and then disable it later in the command line. It is therefore not quite equivalent. But it seems to work well enough for me fom a small amount of testing. If you are really set on not having a special option for it, I can live with it for now. I'm also not convinced that my patch entirely removes the LTO stuff in a consistent way.
Quoted 3 lines
Yeah, I really don't want to go down the path of overriding CONFIG options via make/environment logic. I'm also open to turning off LTO on sandbox and on with qemu-* so it gets wider CI testing.
Quoted 5 lines
Yes you did mention that, but the problem is that LTO is very handy with sandbox, to test the strange things that happen. For example, I found the bug where LTO was dropping a linker-list item, using sandbox. We could perhaps make one of the sandbox builds not use LTO, e.g.
sandbox_flattree?Well, the big issue with LTO+sandbox is that it slows down your workflow, so I would think you want the inverse, one platform does enable it?
Quoted 7 lines
Quoted 2 lines
It slows down all boards that use it, actually, so I normally don't want it enabled for my IDE. This is not sandbox-specific.
The degree of slowdown depends on what you're building on. I see several seconds of link time on my laptop and nothing on my desktop (which isn't that high end either, it's not an EPYC or anything) when LTO is/isn't enabled.
For me, on 16-core AMD desktop, incremental build
LTO: real 0m7.744s
No LTO: real 0m3.172s
Quoted 7 lines
On 4-core laptop:
No LTO: real 0m2.429s
LTO: real 0m15.650sPerhaps the disconnect here is that you just don't see any difference? What times are you seeing?
My 16 core desktop is 14 seconds without, 15 seconds with (full build, not just the link time) and I don't recall what my laptop was but since it's longer than the desktop I'm always just ssh'd in there anyhow.
Signature · 1 line
Tom
Re: [PATCH v2 02/13] Makefile: Allow LTO to be disabled for a build
Hi Tom,
Quoted 98 lines · Tom Rini
Quoted 96 lines · Simon Glass
Hi Tom,
Quoted 75 lines · Tom Rini
Quoted 73 lines · Simon Glass
Hi Tom,
Quoted 69 lines · Tom Rini
Quoted 58 lines · Simon Glass
Hi Tom,
Quoted 51 lines · Tom Rini
Quoted 27 lines · Simon Glass
Hi Tom,
Quoted 16 lines · Tom Rini
Quoted 13 lines · Simon Glass
LTO (Link-Time Optimisation) is an very useful feature which can significantly reduce the size of U-Boot binaries. So far it has been made available for selected ARM boards and sandbox.
However, incremental builds are much slower when LTO is used. For example, an incremental build of sandbox takes 2.1 seconds on my machine, but 6.7 seconds with LTO enabled.
Add a LTO_BUILD=n parameter to the build, so it can be disabled during development if needed, for faster builds.
Add some documentation about LTO while we are here.
Quoted 7 lines
Quoted 1 line
Signed-off-by Simon Glass <[email protected]>We don't need this since you can do:
make EXTRA_CFLAGS="-fno-lto" EXTRA_LDFLAGS="-fno-lto"
to pass -fno-lto to compile/linking and disable lto and per
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46905 this has been working for some time.Quoted 4 lines
Thanks for that, it is a big pain point for me, picking up this patch for every series I write. The incremental build time for sandbox goes from 3 seconds to 27 seconds on my laptop with LTO, which is intolerable.
Yeah, I noticed it's visible on my laptop, but not at all on my desktop (i7 vs Ryzen 7).
Quoted 2 lines
The
EXTRA_CFLAGSsays it is for 'Backward compatibility' and it still does the various LTO things (i.e. it changes the build logic). ItWe're unlikely to move to newer Linux kernel kbuild logic so this isn't going away, and there's not much in the way of logic that's changed for LTO that I see.
Quoted 6 lines
seems odd to me to enable the option and then disable it later in the command line. It is therefore not quite equivalent. But it seems to work well enough for me fom a small amount of testing. If you are really set on not having a special option for it, I can live with it for now. I'm also not convinced that my patch entirely removes the LTO stuff in a consistent way.
Quoted 3 lines
Yeah, I really don't want to go down the path of overriding CONFIG options via make/environment logic. I'm also open to turning off LTO on sandbox and on with qemu-* so it gets wider CI testing.
Quoted 5 lines
Yes you did mention that, but the problem is that LTO is very handy with sandbox, to test the strange things that happen. For example, I found the bug where LTO was dropping a linker-list item, using sandbox. We could perhaps make one of the sandbox builds not use LTO, e.g.
sandbox_flattree?Well, the big issue with LTO+sandbox is that it slows down your workflow, so I would think you want the inverse, one platform does enable it?
Quoted 7 lines
Quoted 2 lines
It slows down all boards that use it, actually, so I normally don't want it enabled for my IDE. This is not sandbox-specific.
The degree of slowdown depends on what you're building on. I see several seconds of link time on my laptop and nothing on my desktop (which isn't that high end either, it's not an EPYC or anything) when LTO is/isn't enabled.
For me, on 16-core AMD desktop, incremental build
LTO: real 0m7.744s
No LTO: real 0m3.172sOn 4-core laptop:
No LTO: real 0m2.429s
LTO: real 0m15.650s
Quoted 6 lines
Quoted 2 lines
Perhaps the disconnect here is that you just don't see any difference? What times are you seeing?
My 16 core desktop is 14 seconds without, 15 seconds with (full build, not just the link time) and I don't recall what my laptop was but since it's longer than the desktop I'm always just ssh'd in there anyhow.
Seems very slow. Are you sure that is an incremental build?
Regards,
Simon
Re: [PATCH v2 02/13] Makefile: Allow LTO to be disabled for a build
Quoted 101 lines · Simon Glass
Hi Tom,
Quoted 98 lines · Tom Rini
Quoted 96 lines · Simon Glass
Hi Tom,
Quoted 75 lines · Tom Rini
Quoted 73 lines · Simon Glass
Hi Tom,
Quoted 69 lines · Tom Rini
Quoted 58 lines · Simon Glass
Hi Tom,
Quoted 51 lines · Tom Rini
Quoted 27 lines · Simon Glass
Hi Tom,
Quoted 16 lines · Tom Rini
Quoted 13 lines · Simon Glass
LTO (Link-Time Optimisation) is an very useful feature which can significantly reduce the size of U-Boot binaries. So far it has been made available for selected ARM boards and sandbox.
However, incremental builds are much slower when LTO is used. For example, an incremental build of sandbox takes 2.1 seconds on my machine, but 6.7 seconds with LTO enabled.
Add a LTO_BUILD=n parameter to the build, so it can be disabled during development if needed, for faster builds.
Add some documentation about LTO while we are here.
Quoted 7 lines
Quoted 1 line
Signed-off-by Simon Glass <[email protected]>We don't need this since you can do:
make EXTRA_CFLAGS="-fno-lto" EXTRA_LDFLAGS="-fno-lto"
to pass -fno-lto to compile/linking and disable lto and per
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46905 this has been working for some time.Quoted 4 lines
Thanks for that, it is a big pain point for me, picking up this patch for every series I write. The incremental build time for sandbox goes from 3 seconds to 27 seconds on my laptop with LTO, which is intolerable.
Yeah, I noticed it's visible on my laptop, but not at all on my desktop (i7 vs Ryzen 7).
Quoted 2 lines
The
EXTRA_CFLAGSsays it is for 'Backward compatibility' and it still does the various LTO things (i.e. it changes the build logic). ItWe're unlikely to move to newer Linux kernel kbuild logic so this isn't going away, and there's not much in the way of logic that's changed for LTO that I see.
Quoted 6 lines
seems odd to me to enable the option and then disable it later in the command line. It is therefore not quite equivalent. But it seems to work well enough for me fom a small amount of testing. If you are really set on not having a special option for it, I can live with it for now. I'm also not convinced that my patch entirely removes the LTO stuff in a consistent way.
Quoted 3 lines
Yeah, I really don't want to go down the path of overriding CONFIG options via make/environment logic. I'm also open to turning off LTO on sandbox and on with qemu-* so it gets wider CI testing.
Quoted 5 lines
Yes you did mention that, but the problem is that LTO is very handy with sandbox, to test the strange things that happen. For example, I found the bug where LTO was dropping a linker-list item, using sandbox. We could perhaps make one of the sandbox builds not use LTO, e.g.
sandbox_flattree?Well, the big issue with LTO+sandbox is that it slows down your workflow, so I would think you want the inverse, one platform does enable it?
Quoted 7 lines
Quoted 2 lines
It slows down all boards that use it, actually, so I normally don't want it enabled for my IDE. This is not sandbox-specific.
The degree of slowdown depends on what you're building on. I see several seconds of link time on my laptop and nothing on my desktop (which isn't that high end either, it's not an EPYC or anything) when LTO is/isn't enabled.
For me, on 16-core AMD desktop, incremental build
LTO: real 0m7.744s
No LTO: real 0m3.172sOn 4-core laptop:
No LTO: real 0m2.429s
LTO: real 0m15.650s
Quoted 8 lines
Quoted 6 lines
Quoted 2 lines
Perhaps the disconnect here is that you just don't see any difference? What times are you seeing?
My 16 core desktop is 14 seconds without, 15 seconds with (full build, not just the link time) and I don't recall what my laptop was but since it's longer than the desktop I'm always just ssh'd in there anyhow.
Seems very slow. Are you sure that is an incremental build?
No, it's a from-scratch build, which should be the worst case for this.
Signature · 1 line
Tom
Re: [PATCH v2 02/13] Makefile: Allow LTO to be disabled for a build
Hi Tom,
Quoted 106 lines · Tom Rini
Quoted 104 lines · Simon Glass
Hi Tom,
Quoted 101 lines · Tom Rini
Quoted 98 lines · Simon Glass
Hi Tom,
Quoted 75 lines · Tom Rini
Quoted 73 lines · Simon Glass
Hi Tom,
Quoted 69 lines · Tom Rini
Quoted 58 lines · Simon Glass
Hi Tom,
Quoted 51 lines · Tom Rini
Quoted 27 lines · Simon Glass
Hi Tom,
Quoted 16 lines · Tom Rini
Quoted 13 lines · Simon Glass
LTO (Link-Time Optimisation) is an very useful feature which can significantly reduce the size of U-Boot binaries. So far it has been made available for selected ARM boards and sandbox.
However, incremental builds are much slower when LTO is used. For example, an incremental build of sandbox takes 2.1 seconds on my machine, but 6.7 seconds with LTO enabled.
Add a LTO_BUILD=n parameter to the build, so it can be disabled during development if needed, for faster builds.
Add some documentation about LTO while we are here.
Quoted 7 lines
Quoted 1 line
Signed-off-by Simon Glass <[email protected]>We don't need this since you can do:
make EXTRA_CFLAGS="-fno-lto" EXTRA_LDFLAGS="-fno-lto"
to pass -fno-lto to compile/linking and disable lto and per
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46905 this has been working for some time.Quoted 4 lines
Thanks for that, it is a big pain point for me, picking up this patch for every series I write. The incremental build time for sandbox goes from 3 seconds to 27 seconds on my laptop with LTO, which is intolerable.
Yeah, I noticed it's visible on my laptop, but not at all on my desktop (i7 vs Ryzen 7).
Quoted 2 lines
The
EXTRA_CFLAGSsays it is for 'Backward compatibility' and it still does the various LTO things (i.e. it changes the build logic). ItWe're unlikely to move to newer Linux kernel kbuild logic so this isn't going away, and there's not much in the way of logic that's changed for LTO that I see.
Quoted 6 lines
seems odd to me to enable the option and then disable it later in the command line. It is therefore not quite equivalent. But it seems to work well enough for me fom a small amount of testing. If you are really set on not having a special option for it, I can live with it for now. I'm also not convinced that my patch entirely removes the LTO stuff in a consistent way.
Quoted 3 lines
Yeah, I really don't want to go down the path of overriding CONFIG options via make/environment logic. I'm also open to turning off LTO on sandbox and on with qemu-* so it gets wider CI testing.
Quoted 5 lines
Yes you did mention that, but the problem is that LTO is very handy with sandbox, to test the strange things that happen. For example, I found the bug where LTO was dropping a linker-list item, using sandbox. We could perhaps make one of the sandbox builds not use LTO, e.g.
sandbox_flattree?Well, the big issue with LTO+sandbox is that it slows down your workflow, so I would think you want the inverse, one platform does enable it?
Quoted 7 lines
Quoted 2 lines
It slows down all boards that use it, actually, so I normally don't want it enabled for my IDE. This is not sandbox-specific.
The degree of slowdown depends on what you're building on. I see several seconds of link time on my laptop and nothing on my desktop (which isn't that high end either, it's not an EPYC or anything) when LTO is/isn't enabled.
For me, on 16-core AMD desktop, incremental build
LTO: real 0m7.744s
No LTO: real 0m3.172sOn 4-core laptop:
No LTO: real 0m2.429s
LTO: real 0m15.650sPerhaps the disconnect here is that you just don't see any difference? What times are you seeing?
Quoted 7 lines
Quoted 5 lines
Quoted 3 lines
My 16 core desktop is 14 seconds without, 15 seconds with (full build, not just the link time) and I don't recall what my laptop was but since it's longer than the desktop I'm always just ssh'd in there anyhow.
Seems very slow. Are you sure that is an incremental build?
No, it's a from-scratch build, which should be the worst case for this.
OK. So I don't do fresh builds very often, mostly they are incremental (for obvious reasons...I am developing things incrementally). So the difference is quite dramatic, as you can see.
Does my problem make more sense now?
Regards,
Simon
Re: [PATCH v2 02/13] Makefile: Allow LTO to be disabled for a build
Quoted 115 lines · Simon Glass
Hi Tom,
Quoted 112 lines · Tom Rini
Quoted 109 lines · Simon Glass
Hi Tom,
Quoted 98 lines · Tom Rini
Quoted 96 lines · Simon Glass
Hi Tom,
Quoted 75 lines · Tom Rini
Quoted 73 lines · Simon Glass
Hi Tom,
Quoted 69 lines · Tom Rini
Quoted 58 lines · Simon Glass
Hi Tom,
Quoted 51 lines · Tom Rini
Quoted 27 lines · Simon Glass
Hi Tom,
Quoted 16 lines · Tom Rini
Quoted 13 lines · Simon Glass
LTO (Link-Time Optimisation) is an very useful feature which can
significantly reduce the size of U-Boot binaries. So far it has been
made available for selected ARM boards and sandbox.However, incremental builds are much slower when LTO is used. For example,
an incremental build of sandbox takes 2.1 seconds on my machine, but 6.7
seconds with LTO enabled.Add a LTO_BUILD=n parameter to the build, so it can be disabled during
development if needed, for faster builds.Add some documentation about LTO while we are here.
Quoted 7 lines
Quoted 1 line
Signed-off-by: Simon Glass <[email protected]>
We don't need this since you can do:
make EXTRA_CFLAGS="-fno-lto" EXTRA_LDFLAGS="-fno-lto"
to pass -fno-lto to compile/linking and disable lto and per
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46905 this has been working for some time.Quoted 4 lines
Thanks for that, it is a big pain point for me, picking up this patch for every series I write. The incremental build time for sandbox goes from 3 seconds to 27 seconds on my laptop with LTO, which is intolerable.
Yeah, I noticed it's visible on my laptop, but not at all on my desktop (i7 vs Ryzen 7).
Quoted 2 lines
The
EXTRA_CFLAGSsays it is for 'Backward compatibility' and it still does the various LTO things (i.e. it changes the build logic). ItWe're unlikely to move to newer Linux kernel kbuild logic so this isn't going away, and there's not much in the way of logic that's changed for LTO that I see.
Quoted 6 lines
seems odd to me to enable the option and then disable it later in the command line. It is therefore not quite equivalent. But it seems to work well enough for me fom a small amount of testing. If you are really set on not having a special option for it, I can live with it for now. I'm also not convinced that my patch entirely removes the LTO stuff in a consistent way.
Quoted 3 lines
Yeah, I really don't want to go down the path of overriding CONFIG options via make/environment logic. I'm also open to turning off LTO on sandbox and on with qemu-* so it gets wider CI testing.
Quoted 5 lines
Yes you did mention that, but the problem is that LTO is very handy with sandbox, to test the strange things that happen. For example, I found the bug where LTO was dropping a linker-list item, using sandbox. We could perhaps make one of the sandbox builds not use LTO, e.g.
sandbox_flattree?Well, the big issue with LTO+sandbox is that it slows down your workflow, so I would think you want the inverse, one platform does enable it?
Quoted 7 lines
Quoted 2 lines
It slows down all boards that use it, actually, so I normally don't want it enabled for my IDE. This is not sandbox-specific.
The degree of slowdown depends on what you're building on. I see several seconds of link time on my laptop and nothing on my desktop (which isn't that high end either, it's not an EPYC or anything) when LTO is/isn't enabled.
For me, on 16-core AMD desktop, incremental build
LTO: real 0m7.744s
No LTO: real 0m3.172sOn 4-core laptop:
No LTO: real 0m2.429s
LTO: real 0m15.650sQuoted 6 lines
Quoted 2 lines
Perhaps the disconnect here is that you just don't see any difference? What times are you seeing?
My 16 core desktop is 14 seconds without, 15 seconds with (full build, not just the link time) and I don't recall what my laptop was but since it's longer than the desktop I'm always just ssh'd in there anyhow.
Seems very slow. Are you sure that is an incremental build?
Quoted 7 lines
Quoted 1 line
No, it's a from-scratch build, which should be the worst case for this.
OK. So I don't do fresh builds very often, mostly they are incremental (for obvious reasons...I am developing things incrementally). So the difference is quite dramatic, as you can see.
Does my problem make more sense now?
Yes, I didn't realize incremental non-LTO link was so quick. But no, I don't want to open the door here to other "make iterative build quicker, override the config" options. Maybe just another note to the development docs and see if that leads to a chorus of people asking for this feature after all.
Signature · 1 line
Tom
PATCH v2 3/13 sandbox: start: Sort the header files
In reply to Simon GlassThese header files don't follow the correct order. Fix this.
---
(no changes since v1)
arch/sandbox/cpu/start.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
arch/sandbox/cpu/start.c +2 -2
@@ -4,14 +4,13 @@*/ #include <common.h> +#include <cli.h> #include <command.h> -#include <dm/root.h> #include <efi_loader.h> #include <errno.h> #include <init.h> #include <log.h> #include <os.h> -#include <cli.h> #include <sort.h> #include <asm/getopt.h> #include <asm/global_data.h>
@@ -19,6 +18,7 @@#include <asm/malloc.h> #include <asm/sections.h> #include <asm/state.h> +#include <dm/root.h> #include <linux/ctype.h> DECLARE_GLOBAL_DATA_PTR ;
Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
Re: [PATCH v2 03/13] sandbox: start: Sort the header files
Quoted 3 lines · Simon Glass
These header files don't follow the correct order. Fix this.
Signed-off-by Simon Glass <[email protected]>
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
PATCH v2 4/13 binman: Expand elf support a little
In reply to Simon GlassAllow finding a symbol by its address. Also export the function to get the file offset of a particular address, so it can be used by a script to be added.
---
(no changes since v1)
tools/binman/elf.py | 58 +++++++++++++++++++++++++++++++++++++++------ 1 file changed, 51 insertions(+), 7 deletions(-)
tools/binman/elf.py +51 -7
@@ -85,6 +85,57 @@ def GetSymbols(fname, patterns): # Sort dict by address return OrderedDict(sorted(syms.items(), key=lambda x: x[1].address)) +def _GetFileOffset(elf, addr): + """Get the file offset for an address + + Args: + elf (ELFFile): ELF file to check + addr (int): Address to search for + + Returns + int: Offset of that address in the ELF file, or None if not valid + """ + 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'] + +def GetFileOffset(fname, addr): + """Get the file offset for an address + + Args: + fname (str): Filename of ELF file to check + addr (int): Address to search for + + Returns + int: Offset of that address in the ELF file, or None if not valid + """ + if not ELF_TOOLS: + raise ValueError('Python elftools package is not available') + with open(fname, 'rb') as fd: + elf = ELFFile(fd) + return _GetFileOffset(elf, addr) + +def GetSymbolFromAddress(fname, addr): + """Get the symbol at a particular address + + Args: + fname (str): Filename of ELF file to check + addr (int): Address to search for + + Returns: + str: Symbol name, or None if no symbol at that address + """ + if not ELF_TOOLS: + raise ValueError('Python elftools package is not available') + with open(fname, 'rb') as fd: + elf = ELFFile(fd) + syms = GetSymbols(fname, None) + for name, sym in syms.items(): + if sym.address == addr: + return name + def GetSymbolFileOffset(fname, patterns): """Get the symbols from an ELF file
@@ -97,13 +148,6 @@ def GetSymbolFileOffset(fname, patterns): 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')
Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
Re: [PATCH v2 04/13] binman: Expand elf support a little
Quoted 5 lines · Simon Glass
Allow finding a symbol by its address. Also export the function to get the file offset of a particular address, so it can be used by a script to be added.
Signed-off-by Simon Glass <[email protected]>
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
PATCH v2 5/13 event: Add basic support for events
In reply to Simon GlassAdd a way to create and dispatch events without needing to allocate memory. Also add a way to 'spy' on events, thus allowing 'hooks' to be created.
Use a linker list for static events, which we can use to replace functions like arch_cpu_init_f(). Allow an EVENT_DEBUG option which makes it easier to see what is going on at runtime, but uses more code space.
Dynamic events allow the creation of a spy at runtime. This is not always necessary, but can be enabled with EVENT_DYNAMIC if needed.
A 'test' event is the only option for now.
---
Changes in v2:
- Add a 'used' attribute to avoid LTO dropping the code
MAINTAINERS | 6 + common/Kconfig | 31 +++++ common/Makefile | 2 + common/board_r.c | 1 + common/event.c | 186 +++++++++++++++++++++++++++++ common/log.c | 1 + include/asm-generic/global_data.h | 13 +++ include/event.h | 188 ++++++++++++++++++++++++++++++ include/event_internal.h | 35 ++++++ include/log.h | 2 + 10 files changed, 465 insertions(+) create mode 100644 common/event.c create mode 100644 include/event.h create mode 100644 include/event_internal.h
MAINTAINERS +6 -0
@@ -809,6 +809,12 @@ S: Maintained F: doc/usage/environment.rst F: scripts/env2string.awk +EVENTS +M: Simon Glass <[email protected]> +S: Maintained +F: common/event.c +F: include/event.h + FASTBOOT S: Orphaned F: cmd/fastboot.c
common/Kconfig +31 -0
@@ -492,6 +492,37 @@ config DISPLAY_BOARDINFO_LATE menu "Start-up hooks" +config EVENT + bool "General-purpose event-handling mechanism" + default y if SANDBOX + help + This enables sending and processing of events, to allow interested + parties to be alerted when something happens. This is an attempt to + step the flow of weak functions, hooks, functions in board_f.c + and board_r.c and the Kconfig options below. + + See doc/develop/event.rst for more information. + +if EVENT + +config EVENT_DYNAMIC + bool "Support event registration at runtime" + default y if SANDBOX + help + Enable this to support adding an event spy at runtime, without adding + it to the EVENT_SPy() linker list. This increases code size slightly + but provides more flexibility for boards and subsystems that need it. + +config EVENT_DEBUG + bool "Enable event debugging assistance" + default y if SANDBOX + help + Enable this get usefui features for seeing what is happening with + events, such as event-type names. This adds to the code size of + U-Boot so can be turned off for production builds. + +endif # EVENT + config ARCH_EARLY_INIT_R bool "Call arch-specific init soon after relocation" help
common/Makefile +2 -0
@@ -89,6 +89,8 @@ obj-y += malloc_simple.o endif endif +obj-$(CONFIG_$(SPL_TPL_)EVENT) += event.o + obj-$(CONFIG_$(SPL_TPL_)HASH) += hash.o obj-$(CONFIG_IO_TRACE) += iotrace.o obj-y += memsize.o
common/board_r.c +1 -0
@@ -594,6 +594,7 @@ static int run_main_loop(void)static init_fnc_t init_sequence_r[] = { initr_trace, initr_reloc, + event_init,/* TODO: could x86/PPC have this also perhaps? */ #if defined(CONFIG_ARM) || defined(CONFIG_RISCV) initr_caches,
common/event.c new file +186 -0
186 lines not on this page · raw message
common/log.c +1 -0
@@ -28,6 +28,7 @@ static const char *const log_cat_name[] = {"devres" ,"acpi" ,"boot" , +"event" , }; _Static_assert(ARRAY_SIZE(log_cat_name) == LOGC_COUNT - LOGC_NONE,
include/asm-generic/global_data.h +13 -0
@@ -20,6 +20,7 @@*/ #ifndef __ASSEMBLY__ +#include <event_internal.h> #include <fdtdec.h> #include <membuff.h> #include <linux/list.h>
@@ -467,6 +468,12 @@ struct global_data {*/ char *smbios_version;#endif +#if CONFIG_IS_ENABLED (EVENT ) +/** +* @event_state: Points to the current state of events +*/ +struct event_state event_state; +#endif };#ifndef DO_DEPS_ONLY static_assert (sizeof (struct global_data ) ==GD_SIZE );
@@ -532,6 +539,12 @@ static_assert(sizeof(struct global_data) == GD_SIZE);#define gd_set_multi_dtb_fit (_dtb)#endif +#if CONFIG_IS_ENABLED (EVENT_DYNAMIC ) +#define gd_event_state () ((struct event_state *)&gd->event_state) +#else +#define gd_event_state () NULL +#endif +/** * enum gd_flags - global data flags *
include/event.h new file +188 -0
188 lines not on this page · raw message
include/event_internal.h new file +35 -0
@@ -0,0 +1,35 @@ +/* SPDX-License-Identifier: GPL-2.0+ */ +/* +* Internal definitions for events +* +* Copyright 2021 Google LLC +* Written by Simon Glass <[email protected]> +*/ + +#ifndef __event_internal_h +#define __event_internal_h + +#include <event.h> +#include <linux/list.h> + +/** +* struct event_spy - a spy that watches for an event of a particular type +* +* @id: Spy ID +* @type: Event type to subscribe to +* @func: Function to call when the event is sent +* @ctx: Context to pass to the function +*/ +struct event_spy { +struct list_head sibling_node; +const char *id; +enum event_t type; +event_handler_t func; +void *ctx; +}; + +struct event_state { +struct list_head spy_head; +}; + +#endif
include/log.h +2 -0
@@ -98,6 +98,8 @@ enum log_category_t {LOGC_ACPI ,/** @LOGC_BOOT: Related to boot process / boot image processing */ LOGC_BOOT , +/** @LOGC_EVENT: Related to event and event handling */ +LOGC_EVENT ,/** @LOGC_COUNT: Number of log categories */ LOGC_COUNT,/** @LOGC_END: Sentinel value for lists of log categories */
Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
Re: [PATCH v2 05/13] event: Add basic support for events
Hi Simon,
Quoted 56 lines · Simon Glass
Add a way to create and dispatch events without needing to allocate memory. Also add a way to 'spy' on events, thus allowing 'hooks' to be created.
Use a linker list for static events, which we can use to replace functions like
arch_cpu_init_f(). Allow anEVENT_DEBUGoption which makes it easier to see what is going on at runtime, but uses more code space.Dynamic events allow the creation of a spy at runtime. This is not always necessary, but can be enabled with
EVENT_DYNAMICif needed.A 'test' event is the only option for now.
Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- Add a 'used' attribute to avoid LTO dropping the codeMAINTAINERS | 6 + common/Kconfig | 31 +++++ common/Makefile | 2 + common/board_r.c | 1 + common/event.c | 186 +++++++++++++++++++++++++++++ common/log.c | 1 + include/asm-generic/global_data.h | 13 +++ include/event.h | 188 ++++++++++++++++++++++++++++++ include/event_internal.h | 35 ++++++ include/log.h | 2 + 10 files changed, 465 insertions(+) create mode 100644 common/event.c create mode 100644 include/event.h create mode 100644 include/event_internal.hMAINTAINERS +6
-0@@ -809,6 +809,12 @@ S: Maintained F: doc/usage/environment.rst F: scripts/env2string.awk +EVENTS +M: Simon Glass <[email protected]> +S: Maintained +F: common/event.c +F: include/event.h + FASTBOOT S: Orphaned F: cmd/fastboot.ccommon/Kconfig +0
-0
common/Kconfig config DISPLAY_BOARDINFO_LATE · line 496
Quoted 5 lines
common/Kconfig +2
-0@@ -492,6 +492,37 @@ config DISPLAY_BOARDINFO_LATE menu "Start-up hooks" +config EVENT + bool "General-purpose event-handling mechanism"
I don't think that this config option needs to be visible (or user-selectable). Instead, any subsystem that needs it should explicitly select it. I prefer that subsystem can add "select EVENT or DM_EVENT" rather than "imply" or "depends on".
-Takahiro Akashi
Quoted 550 lines
common/Kconfig +29
-0+ default y if SANDBOX + help + This enables sending and processing of events, to allow interested + parties to be alerted when something happens. This is an attempt to + step the flow of weak functions, hooks, functions in board_f.c + and board_r.c and the Kconfig options below. + + See doc/develop/event.rst for more information. + +if EVENT + +config EVENT_DYNAMIC + bool "Support event registration at runtime" + default y if SANDBOX + help + Enable this to support adding an event spy at runtime, without adding + it to the EVENT_SPy() linker list. This increases code size slightly + but provides more flexibility for boards and subsystems that need it. + +config EVENT_DEBUG + bool "Enable event debugging assistance" + default y if SANDBOX + help + Enable this get usefui features for seeing what is happening with + events, such as event-type names. This adds to the code size of + U-Boot so can be turned off for production builds. + +endif # EVENT + config ARCH_EARLY_INIT_R bool "Call arch-specific init soon after relocation" helpcommon/Makefile +2
-0@@ -89,6 +89,8 @@ obj-y += malloc_simple.o endif endif +obj-$(CONFIG_$(SPL_TPL_)EVENT) += event.o + obj-$(CONFIG_$(SPL_TPL_)HASH) += hash.o obj-$(CONFIG_IO_TRACE) += iotrace.o obj-y += memsize.ocommon/board_r.c +1
-0@@ -594,6 +594,7 @@ static int run_main_loop(void)static init_fnc_t init_sequence_r[] = { initr_trace, initr_reloc, + event_init,/* TODO: could x86/PPC have this also perhaps? */ #if defined(CONFIG_ARM) || defined(CONFIG_RISCV) initr_caches,common/event.c new file +186
-0@@ -0,0 +1,186 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* Events provide a general-purpose way to react to / subscribe to changes +* within U-Boot +* +* Copyright 2021 Google LLC +* Written by Simon Glass <[email protected]> +*/ + +#define LOG_CATEGORY LOGC_EVENT + +#include <common.h> +#include <event.h> +#include <event_internal.h> +#include <log.h> +#include <linker_lists.h> +#include <malloc.h> +#include <asm/global_data.h> +#include <linux/list.h> + +DECLARE_GLOBAL_DATA_PTR ; + +#if CONFIG_IS_ENABLED (EVENT_DEBUG ) +const char *const type_name[] = { +"none" , +"test" , +}; + +_Static_assert (ARRAY_SIZE (type_name) ==EVT_COUNT ,"event type_name size" ); +#endif + +static const char *event_type_name (enum event_t type) +{ +#if CONFIG_IS_ENABLED (EVENT_DEBUG ) +return type_name[type]; +#else +return "(unknown)" ; +#endif +} + +static int notify_static (struct event *ev) +{ +struct evspy_info *start = + ll_entry_start(struct evspy_info, evspy_info); + const int n_ents = ll_entry_count(struct evspy_info, evspy_info); +struct evspy_info *spy; + +for (spy = start; spy != start + n_ents; spy++) { +if (spy->type == ev->type) { +int ret; + +log_debug ("Sending event %x/%s to spy '%s'\n" , ev->type, +event_type_name (ev->type),event_spy_id (spy)); + ret = spy->func (NULL , ev); + +/* +* TODO: Handle various return codes to +* +* - claim an event (no others will see it) +* - return an error from the event +*/ +if (ret) +return log_msg_ret ("spy" , ret); + } + } + +return 0 ; +} + +static int notify_dynamic (struct event *ev) +{ +struct event_state *state =gd_event_state (); +struct event_spy *spy, *next; + +list_for_each_entry_safe (spy, next, &state->spy_head, sibling_node) { +if (spy->type == ev->type) { +int ret; + +log_debug ("Sending event %x/%s to spy '%s'\n" , ev->type, +event_type_name (ev->type), spy->id); + ret = spy->func (spy->ctx, ev); + +/* +* TODO: Handle various return codes to +* +* - claim an event (no others will see it) +* - return an error from the event +*/ +if (ret) +return log_msg_ret ("spy" , ret); + } + } + +return 0 ; +} + +int event_notify (enum event_t type,void *data,int size) +{ +struct event event; +int ret; + + event.type = type; +if (size >sizeof (event.data)) +return log_msg_ret ("size" , -E2BIG ); +memcpy (&event.data, data, size); + + ret =notify_static (&event); +if (ret) +return log_msg_ret ("dyn" , ret); + +if (CONFIG_IS_ENABLED (EVENT_DYNAMIC )) { + ret =notify_dynamic (&event); +if (ret) +return log_msg_ret ("dyn" , ret); + } + +return 0 ; +} + +int event_notify_null (enum event_t type) +{ +return event_notify (type,NULL ,0 ); +} + +void event_show_spy_list (void ) +{ +struct evspy_info *start = + ll_entry_start(struct evspy_info, evspy_info); + const int n_ents = ll_entry_count(struct evspy_info, evspy_info); +struct evspy_info *spy; +const int size =sizeof (ulong) *2 ; + +printf ("Seq %-24s %*s %s\n" ,"Type" , size,"Function" ,"ID" ); +for (spy = start; spy != start + n_ents; spy++) { +printf ("%3x %-3x %-20s %*p %s\n" , (uint)(spy - start), + spy->type,event_type_name (spy->type), size, spy->func, +event_spy_id (spy)); + } +} + +#if CONFIG_IS_ENABLED (EVENT_DYNAMIC ) +static void spy_free (struct event_spy *spy) +{ +list_del (&spy->sibling_node); +} + +int event_register (const char *id,enum event_t type,event_handler_t func,void *ctx) +{ +struct event_state *state =gd_event_state (); +struct event_spy *spy; + +if (!CONFIG_IS_ENABLED (EVENT_DYNAMIC )) +return -ENOSYS ; + spy =malloc (sizeof (*spy)); +if (!spy) +return log_msg_ret ("alloc" , -ENOMEM ); + + spy->id = id; + spy->type = type; + spy->func = func; + spy->ctx = ctx; +list_add_tail (&spy->sibling_node, &state->spy_head); + +return 0 ; +} + +int event_uninit (void ) +{ +struct event_state *state =gd_event_state (); +struct event_spy *spy, *next; + +list_for_each_entry_safe (spy, next, &state->spy_head, sibling_node) +spy_free (spy); + +return 0 ; +} + +int event_init (void ) +{ +struct event_state *state =gd_event_state (); + +INIT_LIST_HEAD (&state->spy_head); + +return 0 ; +} +#endif /* EVENT_DYNAMIC */ common/log.c +1
-0@@ -28,6 +28,7 @@ static const char *const log_cat_name[] = {"devres" ,"acpi" ,"boot" , +"event" , }; _Static_assert(ARRAY_SIZE(log_cat_name) == LOGC_COUNT - LOGC_NONE,include/asm-generic/global_data.h +13
-0@@ -20,6 +20,7 @@*/ #ifndef __ASSEMBLY__ +#include <event_internal.h> #include <fdtdec.h> #include <membuff.h> #include <linux/list.h> @@ -467,6 +468,12 @@ struct global_data {*/ char *smbios_version;#endif +#if CONFIG_IS_ENABLED (EVENT ) +/** +* @event_state: Points to the current state of events +*/ +struct event_state event_state; +#endif };#ifndef DO_DEPS_ONLY static_assert (sizeof (struct global_data ) ==GD_SIZE );@@ -532,6 +539,12 @@ static_assert(sizeof(struct global_data) == GD_SIZE);#define gd_set_multi_dtb_fit (_dtb)#endif +#if CONFIG_IS_ENABLED (EVENT_DYNAMIC ) +#define gd_event_state () ((struct event_state *)&gd->event_state) +#else +#define gd_event_state () NULL +#endif +/** * enum gd_flags - global data flags * include/event.h new file +188
-0@@ -0,0 +1,188 @@ +/* SPDX-License-Identifier: GPL-2.0+ */ +/* +* Events provide a general-purpose way to react to / subscribe to changes +* within U-Boot +* +* Copyright 2021 Google LLC +* Written by Simon Glass <[email protected]> +*/ + +#ifndef __event_h +#define __event_h + +/** +* enum event_t - Types of events supported by U-Boot +* +* @EVT_DM_PRE_PROBE: Device is about to be probed +*/ +enum event_t { +EVT_NONE , +EVT_TEST , + +EVT_COUNT +}; + +union event_data { +/** +* struct event_data_test - test data +* +* @signal: A value to update the state with +*/ +struct event_data_test { +int signal; + } test; +}; + +/** +* struct event - an event that can be sent and received +* +* @type: Event type +* @data: Data for this particular event +*/ +struct event { +enum event_t type; +union event_data data; +}; + +/** Function type for event handlers */ +typedef int (*event_handler_t )(void *ctx,struct event *event); + +/** +* struct evspy_info - information about an event spy +* +* @func: Function to call when the event is activated (must be first) +* @type: Event type +* @id: Event id string +*/ +struct evspy_info { +event_handler_t func; +enum event_t type; +#if CONFIG_IS_ENABLED (EVENT_DEBUG ) +const char *id; +#endif +}; + +/* Declare a new event spy */ +#if CONFIG_IS_ENABLED (EVENT_DEBUG ) +#define _ESPY_REC (_type, _func) { _func, _type, #_func, } +#else +#define _ESPY_REC (_type, _func) { _func, _type, } +#endif + +static inline const char *event_spy_id (struct evspy_info *spy) +{ +#if CONFIG_IS_ENABLED (EVENT_DEBUG ) +return spy->id; +#else +return "?" ; +#endif +} + +/* +* It seems that LTO will drop list entries if it decides they are not used, +* although the conditions that cause this are unclear. +* +* The example found is the following: +* +* static int sandbox_misc_init_f(void *ctx, struct event *event) +* { +* return sandbox_early_getopt_check(); +* } +* EVENT_SPY(EVT_MISC_INIT_F, sandbox_misc_init_f); +* +* where EVENT_SPY uses ll_entry_declare() +* +* In this case, LTO decides to drop the sandbox_misc_init_f() function +* (which is fine) but then drops the linker-list entry too. This means +* that the code no longer works, in this case sandbox no-longer checks its +* command-line arguments properly. +* +* Without LTO, the KEEP() command in the .lds file is enough to keep the +* entry around. But with LTO it seems that the entry has already been +* dropped before the link script is considered. +* +* The only solution I can think of is to mark linker-list entries as 'used' +* using an attribute. This should be safe, since we don't actually want to drop +* any of these. However this does slightly limit LTO's optimisation choices. +*/ +#define EVENT_SPY (_type, _func) \ + static __attribute__((used)) ll_entry_declare(struct evspy_info, \ + _type, evspy_info) = \ + _ESPY_REC(_type, _func) + +/** +* event_register - register a new spy +* +* @id: Spy ID +* @type: Event type to subscribe to +* @func: Function to call when the event is sent +* @ctx: Context to pass to the function +* @return 0 if OK, -ve on error +*/ +int event_register (const char *id,enum event_t type,event_handler_t func, +void *ctx); + +#if CONFIG_IS_ENABLED (EVENT ) +/** +* event_notify() - notify spies about an event +* +* It is possible to pass in union event_data here but that may not be +* convenient if the data is elsewhere, or is one of the members of the union. +* So this uses a void * for @data, with a separate @size. +* +* @type: Event type +* @data: Event data to be sent (e.g. union_event_data) +* @size: Size of data in bytes +* @return 0 if OK, -ve on error +*/ +int event_notify (enum event_t type,void *data,int size); + +/** +* event_notify_null() - notify spies about an event +* +* Data is NULL and the size is 0 +* +* @type: Event type +* @return 0 if OK, -ve on error +*/ +int event_notify_null (enum event_t type); +#else +static inline int event_notify (enum event_t type,void *data,int size) +{ +return 0 ; +} + +static inline int event_notify_null (enum event_t type) +{ +return 0 ; +} +#endif + +#if CONFIG_IS_ENABLED (EVENT_DYNAMIC ) +/** +* event_uninit() - Clean up dynamic events +* +* This removes all dynamic event handlers +*/ +int event_uninit (void ); + +/** +* event_uninit() - Set up dynamic events +* +* Init a list of dynamic event handlers, so that these can be added as +* needed +*/ +int event_init (void ); +#else +static inline int event_uninit (void ) +{ +return 0 ; +} + +static inline int event_init (void ) +{ +return 0 ; +} +#endif + +#endif include/event_internal.h new file +35
-0@@ -0,0 +1,35 @@ +/* SPDX-License-Identifier: GPL-2.0+ */ +/* +* Internal definitions for events +* +* Copyright 2021 Google LLC +* Written by Simon Glass <[email protected]> +*/ + +#ifndef __event_internal_h +#define __event_internal_h + +#include <event.h> +#include <linux/list.h> + +/** +* struct event_spy - a spy that watches for an event of a particular type +* +* @id: Spy ID +* @type: Event type to subscribe to +* @func: Function to call when the event is sent +* @ctx: Context to pass to the function +*/ +struct event_spy { +struct list_head sibling_node; +const char *id; +enum event_t type; +event_handler_t func; +void *ctx; +}; + +struct event_state { +struct list_head spy_head; +}; + +#endif include/log.h +2
-0@@ -98,6 +98,8 @@ enum log_category_t {LOGC_ACPI ,/** @LOGC_BOOT: Related to boot process / boot image processing */ LOGC_BOOT , +/** @LOGC_EVENT: Related to event and event handling */ +LOGC_EVENT ,/** @LOGC_COUNT: Number of log categories */ LOGC_COUNT,/** @LOGC_END: Sentinel value for lists of log categories */ Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
Re: [PATCH v2 05/13] event: Add basic support for events
Hi Takahiro,
Quoted 63 lines · AKASHI Takahiro
Hi Simon,
Quoted 59 lines · Simon Glass
Add a way to create and dispatch events without needing to allocate memory. Also add a way to 'spy' on events, thus allowing 'hooks' to be created.
Use a linker list for static events, which we can use to replace functions like
arch_cpu_init_f(). Allow anEVENT_DEBUGoption which makes it easier to see what is going on at runtime, but uses more code space.Dynamic events allow the creation of a spy at runtime. This is not always necessary, but can be enabled with
EVENT_DYNAMICif needed.A 'test' event is the only option for now.
Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- Add a 'used' attribute to avoid LTO dropping the codeMAINTAINERS | 6 + common/Kconfig | 31 +++++ common/Makefile | 2 + common/board_r.c | 1 + common/event.c | 186 +++++++++++++++++++++++++++++ common/log.c | 1 + include/asm-generic/global_data.h | 13 +++ include/event.h | 188 ++++++++++++++++++++++++++++++ include/event_internal.h | 35 ++++++ include/log.h | 2 + 10 files changed, 465 insertions(+) create mode 100644 common/event.c create mode 100644 include/event.h create mode 100644 include/event_internal.hMAINTAINERS +6
-0@@ -809,6 +809,12 @@ S: Maintained F: doc/usage/environment.rst F: scripts/env2string.awk +EVENTS +M: Simon Glass <[email protected]> +S: Maintained +F: common/event.c +F: include/event.h + FASTBOOT S: Orphaned F: cmd/fastboot.ccommon/Kconfig +0
-0@@ -492,6 +492,37 @@ config DISPLAY_BOARDINFO_LATE menu "Start-up hooks"
Quoted 7 lines
common/Kconfig config DISPLAY_BOARDINFO_LATE · line 496
Quoted 2 lines
common/Kconfig +2
-0+config EVENT + bool "General-purpose event-handling mechanism"I don't think that this config option needs to be visible (or user-selectable). Instead, any subsystem that needs it should explicitly select it. I prefer that subsystem can add "select EVENT or
DM_EVENT" rather than "imply" or "depends on".
But events can be added for other reasons. For example a board may want to detect that a USB controller is about to be probed and enable power to devices on that USB bus so that the devices are enumerated. This series is partly about replacing all the various hooks we currently have but goes further than that.
[..]
Regards,
Simon
Re: [PATCH v2 05/13] event: Add basic support for events
In reply to Simon GlassQuoted 8 lines · Simon Glass
Add a way to create and dispatch events without needing to allocate memory. Also add a way to 'spy' on events, thus allowing 'hooks' to be created.
Use a linker list for static events, which we can use to replace functions like
arch_cpu_init_f(). Allow anEVENT_DEBUGoption which makes it easier to see what is going on at runtime, but uses more code space.
Quoted 6 lines
Dynamic events allow the creation of a spy at runtime. This is not always necessary, but can be enabled with
EVENT_DYNAMICif needed.A 'test' event is the only option for now.
Signed-off-by Simon Glass <[email protected]>
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
PATCH v2 6/13 event: Add a simple test
In reply to Simon GlassAdd a test for event registration and activation.
---
(no changes since v1)
MAINTAINERS | 1 + test/common/Makefile | 1 + test/common/event.c | 47 ++++++++++++++++++++++++++++++++++++++++++++ test/test-main.c | 4 ++++ 4 files changed, 53 insertions(+) create mode 100644 test/common/event.c
MAINTAINERS +1 -0
@@ -814,6 +814,7 @@ M: Simon Glass <[email protected]> S: Maintained F: common/event.c F: include/event.h +F: test/common/event.c FASTBOOT S: Orphaned
test/common/Makefile +1 -0
@@ -1,3 +1,4 @@ # SPDX-License-Identifier: GPL-2.0+ obj-y += cmd_ut_common.o obj-$(CONFIG_AUTOBOOT) += test_autoboot.o +obj-$(CONFIG_EVENT) += event.o
test/common/event.c new file +47 -0
47 lines not on this page · raw message
test/test-main.c +4 -0
@@ -7,6 +7,7 @@#include <common.h> #include <console.h> #include <dm.h> +#include <event.h> #include <dm/root.h> #include <dm/test.h> #include <dm/uclass-internal.h>
@@ -218,6 +219,8 @@ static int dm_test_restore(struct device_node *of_root)*/ static int test_pre_run (struct unit_test_state *uts,struct unit_test *test) { +ut_assertok (event_init ()); +if (test->flags &UT_TESTF_DM )ut_assertok (dm_test_pre_run (uts));
@@ -260,6 +263,7 @@ static int test_post_run(struct unit_test_state *uts, struct unit_test *test)ut_unsilence_console (uts);if (test->flags &UT_TESTF_DM )ut_assertok (dm_test_post_run (uts)); +ut_assertok (event_uninit ());return 0 ; }
Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
Re: [PATCH v2 06/13] event: Add a simple test
Quoted 3 lines · Simon Glass
Add a test for event registration and activation.
Signed-off-by Simon Glass <[email protected]>
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
PATCH v2 7/13 event: Set up the event system on start-up
In reply to Simon GlassCall event_init() before relocation to get the event system running.
---
(no changes since v1)
common/board_f.c | 2 ++ 1 file changed, 2 insertions(+)
common/board_f.c +2 -0
@@ -19,6 +19,7 @@#include <dm.h> #include <env.h> #include <env_internal.h> +#include <event.h> #include <fdtdec.h> #include <fs.h> #include <hang.h>
@@ -828,6 +829,7 @@ static const init_fnc_t init_sequence_f[] = { initf_malloc, log_init, initf_bootstage,/* uses its own timer, so does not need DM */ + event_init, #ifdef CONFIG_BLOBLIST bloblist_init, #endif
Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
Re: [PATCH v2 07/13] event: Set up the event system on start-up
Quoted 3 lines · Simon Glass
Call
event_init()before relocation to get the event system running.Signed-off-by Simon Glass <[email protected]>
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
PATCH v2 8/13 event: Add events for device probe/remove
In reply to Simon GlassGenerate events when devices are probed or removed, allowing hooks to be added for these situations.
This is controlled by the DM_EVENT config option.
---
(no changes since v1)
common/event.c | 6 ++++++ drivers/core/Kconfig | 10 ++++++++++ drivers/core/device-remove.c | 8 ++++++++ drivers/core/device.c | 9 +++++++++ include/dm/device-internal.h | 10 ++++++++++ include/event.h | 15 ++++++++++++++ test/common/event.c | 38 ++++++++++++++++++++++++++++++++++++ 7 files changed, 96 insertions(+)
common/event.c +6 -0
@@ -24,6 +24,12 @@ DECLARE_GLOBAL_DATA_PTR;const char *const type_name[] = {"none" ,"test" , + +/* Events related to driver model */ +"dm_pre_probe" , +"dm_post_probe" , +"dm_pre_remove" , +"dm_post_remove" , };_Static_assert (ARRAY_SIZE (type_name) ==EVT_COUNT ,"event type_name size" );
drivers/core/Kconfig +10 -0
@@ -77,6 +77,16 @@ config DM_DEVICE_REMOVE it causes unplugged devices to linger around in the dm-tree, and it causes USB host controllers to not be stopped when booting the OS. +config DM_EVENT + bool "Support events with driver model" + depends on DM + imply EVENT + default y if SANDBOX + help + This enables support for generating events related to driver model + operations, such as prbing or removing a device. Subsystems can + register a 'spy' function that is called when the event occurs. + config SPL_DM_DEVICE_REMOVE bool "Support device removal in SPL" depends on SPL_DM
drivers/core/device-remove.c +8 -0
@@ -207,6 +207,10 @@ int device_remove(struct udevice *dev, uint flags)if (!(dev_get_flags (dev) &DM_FLAG_ACTIVATED ))return 0 ; + ret =device_notify (dev,EVT_DM_PRE_REMOVE ); +if (ret) +return ret; +/* * If the child returns EKEYREJECTED, continue. It just means that it * didn't match the flags.
@@ -256,6 +260,10 @@ int device_remove(struct udevice *dev, uint flags)dev_bic_flags (dev,DM_FLAG_ACTIVATED ); + ret =device_notify (dev,EVT_DM_POST_REMOVE ); +if (ret) + goto err_remove; +return 0 ; err_remove:
drivers/core/device.c +9 -0
@@ -10,6 +10,7 @@#include <common.h> #include <cpu_func.h> +#include <event.h> #include <log.h> #include <asm/global_data.h> #include <asm/io.h>
@@ -493,6 +494,10 @@ int device_probe(struct udevice *dev)if (dev_get_flags (dev) &DM_FLAG_ACTIVATED )return 0 ; + ret =device_notify (dev,EVT_DM_PRE_PROBE ); +if (ret) +return ret; + drv = dev->driver;assert (drv);
@@ -597,6 +602,10 @@ int device_probe(struct udevice *dev) dev->name, ret, errno_str(ret)); } + ret =device_notify (dev,EVT_DM_POST_PROBE ); +if (ret) +return ret; +return 0 ; fail_uclass:if (device_remove (dev,DM_REMOVE_NORMAL )) {
include/dm/device-internal.h +10 -0
@@ -10,6 +10,7 @@#ifndef _DM_DEVICE_INTERNAL_H#define _DM_DEVICE_INTERNAL_H +#include <event.h> #include <linker_lists.h> #include <dm/ofnode.h>
@@ -426,4 +427,13 @@ static inline void devres_release_all(struct udevice *dev) }#endif /* ! CONFIG_DEVRES */ + +static inline int device_notify (const struct udevice *dev,enum event_t type) +{ +#if CONFIG_IS_ENABLED (DM_EVENT ) +return event_notify (type, &dev,sizeof (dev)); +#else +return 0 ; +#endif +}#endif
include/event.h +15 -0
@@ -19,6 +19,12 @@ enum event_t {EVT_NONE ,EVT_TEST , +/* Events related to driver model */ +EVT_DM_PRE_PROBE , +EVT_DM_POST_PROBE , +EVT_DM_PRE_REMOVE , +EVT_DM_POST_REMOVE , + EVT_COUNT };
@@ -31,6 +37,15 @@ union event_data {struct event_data_test {int signal; } test; + +/** +* struct event_dm - driver model event +* +* @dev: Device this event relates to +*/ +struct event_dm { +struct udevice *dev; + } dm; };/**
test/common/event.c +38 -0
@@ -45,3 +45,41 @@ static int test_event_base(struct unit_test_state *uts)return 0 ; }COMMON_TEST (test_event_base,0 ); + +static int h_probe (void *ctx,struct event *event) +{ +struct test_state *test_state = ctx; + + test_state->dev = event->data.dm.dev; +switch (event->type) { +case EVT_DM_PRE_PROBE : + test_state->val |=1 ; +break ; +case EVT_DM_POST_PROBE : + test_state->val |=2 ; +break ; +default : +break ; + } + +return 0 ; +} + +static int test_event_probe (struct unit_test_state *uts) +{ +struct test_state state; +struct udevice *dev; + + state.val =0 ; +ut_assertok (event_register ("pre" ,EVT_DM_PRE_PROBE , h_probe, &state)); +ut_assertok (event_register ("post" ,EVT_DM_POST_PROBE , h_probe, &state)); + +/* Probe a device */ +ut_assertok (uclass_first_device_err (UCLASS_TEST_FDT , &dev)); + +/* Check that the handler is called */ +ut_asserteq (3 , state.val); + +return 0 ; +} +COMMON_TEST (test_event_probe,UT_TESTF_DM |UT_TESTF_SCAN_FDT );
Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
Re: [PATCH v2 08/13] event: Add events for device probe/remove
Quoted 6 lines · Simon Glass
Generate events when devices are probed or removed, allowing hooks to be added for these situations.
This is controlled by the
DM_EVENTconfig option.Signed-off-by Simon Glass <[email protected]>
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
PATCH v2 9/13 event: Convert misc_init_f() to use events
In reply to Simon GlassThis hook can be implmented using events, for the three boards that actually use it.
Add the event type and event handlers. Drop CONFIG_MISC_INIT_F since we can just use CONFIG_EVENT to control this. Since sandbox always enables CONFIG_EVENT, we can drop the defconfig lines there too.
---
Changes in v2:
- Update keymile pg-wcom boards also
- Make the sandbox function static
arch/sandbox/cpu/start.c | 4 +++- board/google/chromebook_coral/coral.c | 7 +++++-- board/keymile/kmcent2/kmcent2.c | 4 +++- board/keymile/pg-wcom-ls102xa/pg-wcom-ls102xa.c | 5 ++++- common/Kconfig | 6 ------ common/board_f.c | 7 +++++-- common/event.c | 3 +++ configs/chromebook_coral_defconfig | 1 + configs/kmcent2_defconfig | 2 +- configs/pg_wcom_expu1_defconfig | 1 + configs/pg_wcom_expu1_update_defconfig | 1 + configs/pg_wcom_seli8_defconfig | 1 + configs/pg_wcom_seli8_update_defconfig | 1 + configs/sandbox64_defconfig | 1 - configs/sandbox_defconfig | 1 - configs/sandbox_flattree_defconfig | 1 - configs/sandbox_spl_defconfig | 1 - configs/tools-only_defconfig | 1 - include/configs/km/pg-wcom-ls102xa.h | 2 -- include/event.h | 3 +++ include/init.h | 1 - 21 files changed, 32 insertions(+), 22 deletions(-)
arch/sandbox/cpu/start.c +3 -1
@@ -8,6 +8,7 @@#include <command.h> #include <efi_loader.h> #include <errno.h> +#include <event.h> #include <init.h> #include <log.h> #include <os.h>
@@ -119,10 +120,11 @@ int sandbox_early_getopt_check(void)os_exit (0 ); } -int misc_init_f (void ) +static int sandbox_misc_init_f (void *ctx,struct event *event) {return sandbox_early_getopt_check (); } +EVENT_SPY (EVT_MISC_INIT_F , sandbox_misc_init_f);static int sandbox_cmdline_cb_help (struct sandbox_state *state,const char *arg) {
board/google/chromebook_coral/coral.c +5 -2
@@ -10,6 +10,7 @@#include <command.h> #include <cros_ec.h> #include <dm.h> +#include <event.h> #include <init.h> #include <log.h> #include <sysinfo.h>
@@ -32,11 +33,12 @@ struct cros_gpio_info {int flags; }; -int misc_init_f (void ) +static int coral_check_ll_boot (void *ctx,struct event *event) {if (!ll_boot_init ()) {printf ("Running as secondary loader" ); -if (gd->arch.coreboot_table) { +if (CONFIG_IS_ENABLED (COREBOOT_SYSINFO ) && + gd->arch.coreboot_table) {int ret; printf(" (found coreboot table at %lx)",
@@ -55,6 +57,7 @@ int misc_init_f(void)return 0 ; } +EVENT_SPY (EVT_MISC_INIT_F , coral_check_ll_boot);int arch_misc_init (void ) {
board/keymile/kmcent2/kmcent2.c +3 -1
@@ -6,6 +6,7 @@* Copyright 2013 Freescale Semiconductor, Inc. */ +#include <event.h> #include <asm/cache.h> #include <asm/fsl_fdt.h> #include <asm/fsl_law.h>
@@ -181,7 +182,7 @@ unsigned long get_serial_clock(unsigned long dummy)return (gd->bus_clk /2 ); } -int misc_init_f (void ) +static int kmcent2_misc_init_f (void *ctx,struct event *event) {/* configure QRIO pis for i2c deblocking */ i2c_deblock_gpio_cfg ();
@@ -209,6 +210,7 @@ int misc_init_f(void)return 0 ; } +EVENT_SPY (EVT_MISC_INIT_F , kmcent2_misc_init_f);#define USED_SRDS_BANK 0#define EXPECTED_SRDS_RFCK SRDS_PLLCR0_RFCK_SEL_100
board/keymile/pg-wcom-ls102xa/pg-wcom-ls102xa.c +4 -1
@@ -4,6 +4,7 @@*/ #include <common.h> +#include <event.h> #include <i2c.h> #include <asm/io.h> #include <asm/arch/immap_ls102xa.h>
@@ -109,12 +110,14 @@ int board_early_init_f(void)return 0 ; } -int misc_init_f (void ) +static int pg_wcom_misc_init_f (void *ctx,struct event *event) {if (IS_ENABLED (CONFIG_PG_WCOM_UBOOT_UPDATE_SUPPORTED ))check_for_uboot_update (); +return 0 ; } +EVENT_SPY (EVT_MISC_INIT_F , pg_wcom_misc_init_f);int board_init (void ) {
common/Kconfig +0 -6
@@ -583,12 +583,6 @@ config LAST_STAGE_INIT U-Boot calls last_stage_init() before the command-line interpreter is started. -config MISC_INIT_F - bool "Execute pre-relocation misc init" - help - Enabling this option calls the 'misc_init_f' function in the init - sequence just before DRAM is inited. - config MISC_INIT_R bool "Execute Misc Init" default y if ARCH_KEYSTONE || ARCH_SUNXI || MPC85xx
common/board_f.c +5 -2
@@ -818,6 +818,11 @@ __weak int clear_bss(void)return 0 ; } +static int misc_init_f (void ) +{ +return event_notify_null (EVT_MISC_INIT_F ); +} +static const init_fnc_t init_sequence_f[] = { setup_mon_len, #ifdef CONFIG_OF_CONTROL
@@ -877,9 +882,7 @@ static const init_fnc_t init_sequence_f[] = { show_board_info,#endif INIT_FUNC_WATCHDOG_INIT -#if defined(CONFIG_MISC_INIT_F) misc_init_f, -#endif INIT_FUNC_WATCHDOG_RESET #if CONFIG_IS_ENABLED(SYS_I2C_LEGACY) init_func_i2c,
common/event.c +3 -0
@@ -30,6 +30,9 @@ const char *const type_name[] = {"dm_post_probe" ,"dm_pre_remove" ,"dm_post_remove" , + +/* init hooks */ +"misc_init_f" , };_Static_assert (ARRAY_SIZE (type_name) ==EVT_COUNT ,"event type_name size" );
configs/chromebook_coral_defconfig +1 -0
@@ -35,6 +35,7 @@ CONFIG_USE_BOOTCOMMAND=y CONFIG_BOOTCOMMAND="tpm init; tpm startup TPM2_SU_CLEAR; read mmc 0:2 100000 0 80; setexpr loader *001004f0; setexpr size *00100518; setexpr blocks $size / 200; read mmc 0:2 100000 80 $blocks; setexpr setup $loader - 1000; setexpr cmdline_ptr $loader - 2000; setexpr.s cmdline *$cmdline_ptr; setexpr cmdline gsub %U \\\\${uuid}; if part uuid mmc 0:2 uuid; then zboot start 100000 0 0 0 $setup cmdline; zboot load; zboot setup; zboot dump; zboot go;fi" CONFIG_SYS_CONSOLE_INFO_QUIET=y CONFIG_DISPLAY_BOARDINFO_LATE=y +CONFIG_EVENT=y CONFIG_LAST_STAGE_INIT=y CONFIG_BLOBLIST=y # CONFIG_TPL_BLOBLIST is not set
configs/kmcent2_defconfig +1 -1
@@ -16,10 +16,10 @@ CONFIG_FIT=y CONFIG_FIT_VERBOSE=y CONFIG_OF_BOARD_SETUP=y CONFIG_OF_STDOUT_VIA_ALIAS=y +CONFIG_EVENT=y CONFIG_BOARD_EARLY_INIT_F=y CONFIG_BOARD_EARLY_INIT_R=y CONFIG_LAST_STAGE_INIT=y -CONFIG_MISC_INIT_F=y CONFIG_HUSH_PARSER=y CONFIG_CMD_DM=y CONFIG_CMD_I2C=y
configs/pg_wcom_expu1_defconfig +1 -0
@@ -35,6 +35,7 @@ CONFIG_AUTOBOOT_STOP_STR=" " CONFIG_USE_BOOTARGS=y CONFIG_BOOTARGS="console=ttyS0,115200 root=/dev/ram0" CONFIG_SILENT_CONSOLE=y +CONFIG_EVENT=y CONFIG_LAST_STAGE_INIT=y CONFIG_MISC_INIT_R=y CONFIG_CMD_IMLS=y
configs/pg_wcom_expu1_update_defconfig +1 -0
@@ -33,6 +33,7 @@ CONFIG_AUTOBOOT_STOP_STR=" " CONFIG_USE_BOOTARGS=y CONFIG_BOOTARGS="console=ttyS0,115200 root=/dev/ram0" CONFIG_SILENT_CONSOLE=y +CONFIG_EVENT=y CONFIG_LAST_STAGE_INIT=y CONFIG_MISC_INIT_R=y CONFIG_CMD_IMLS=y
configs/pg_wcom_seli8_defconfig +1 -0
@@ -35,6 +35,7 @@ CONFIG_AUTOBOOT_STOP_STR=" " CONFIG_USE_BOOTARGS=y CONFIG_BOOTARGS="console=ttyS0,115200 root=/dev/ram0" CONFIG_SILENT_CONSOLE=y +CONFIG_EVENT=y CONFIG_LAST_STAGE_INIT=y CONFIG_MISC_INIT_R=y CONFIG_CMD_IMLS=y
configs/pg_wcom_seli8_update_defconfig +1 -0
@@ -33,6 +33,7 @@ CONFIG_AUTOBOOT_STOP_STR=" " CONFIG_USE_BOOTARGS=y CONFIG_BOOTARGS="console=ttyS0,115200 root=/dev/ram0" CONFIG_SILENT_CONSOLE=y +CONFIG_EVENT=y CONFIG_LAST_STAGE_INIT=y CONFIG_MISC_INIT_R=y CONFIG_CMD_IMLS=y
configs/sandbox64_defconfig +0 -1
@@ -23,7 +23,6 @@ CONFIG_CONSOLE_RECORD=y CONFIG_CONSOLE_RECORD_OUT_SIZE=0x1000 CONFIG_PRE_CONSOLE_BUFFER=y CONFIG_DISPLAY_BOARDINFO_LATE=y -CONFIG_MISC_INIT_F=y CONFIG_CMD_CPU=y CONFIG_CMD_LICENSE=y CONFIG_CMD_BOOTZ=y
configs/sandbox_defconfig +0 -1
@@ -32,7 +32,6 @@ CONFIG_CONSOLE_RECORD_OUT_SIZE=0x1000 CONFIG_PRE_CONSOLE_BUFFER=y CONFIG_LOG=y CONFIG_DISPLAY_BOARDINFO_LATE=y -CONFIG_MISC_INIT_F=y CONFIG_STACKPROTECTOR=y CONFIG_ANDROID_AB=y CONFIG_CMD_CPU=y
configs/sandbox_flattree_defconfig +0 -1
@@ -20,7 +20,6 @@ CONFIG_BOOTSTAGE_STASH_SIZE=0x4096 CONFIG_CONSOLE_RECORD=y CONFIG_CONSOLE_RECORD_OUT_SIZE=0x1000 CONFIG_DISPLAY_BOARDINFO_LATE=y -CONFIG_MISC_INIT_F=y CONFIG_CMD_CPU=y CONFIG_CMD_LICENSE=y CONFIG_CMD_BOOTZ=y
configs/sandbox_spl_defconfig +0 -1
@@ -30,7 +30,6 @@ CONFIG_BOOTSTAGE_STASH_SIZE=0x4096 CONFIG_CONSOLE_RECORD=y CONFIG_CONSOLE_RECORD_OUT_SIZE=0x1000 CONFIG_DISPLAY_BOARDINFO_LATE=y -CONFIG_MISC_INIT_F=y CONFIG_HANDOFF=y CONFIG_SPL_BOARD_INIT=y CONFIG_SPL_ENV_SUPPORT=y
configs/tools-only_defconfig +0 -1
@@ -9,7 +9,6 @@ CONFIG_TIMESTAMP=y CONFIG_FIT_SIGNATURE=y CONFIG_USE_BOOTCOMMAND=y CONFIG_BOOTCOMMAND="run distro_bootcmd" -CONFIG_MISC_INIT_F=y # CONFIG_CMD_BOOTD is not set # CONFIG_CMD_BOOTM is not set # CONFIG_CMD_ELF is not set
include/configs/km/pg-wcom-ls102xa.h +0 -2
@@ -274,6 +274,4 @@#define CONFIG_SYS_BOOTM_LEN (64 << 20)/* Increase max gunzip size */ #define CONFIG_SYS_BOOTMAPSZ (256 << 20)/* Increase map for Linux */ -#define CONFIG_MISC_INIT_F -#endif
include/event.h +3 -0
@@ -25,6 +25,9 @@ enum event_t {EVT_DM_PRE_REMOVE ,EVT_DM_POST_REMOVE , +/* Init hooks */ +EVT_MISC_INIT_F , + EVT_COUNT };
include/init.h +0 -1
@@ -217,7 +217,6 @@ int init_cache_f_r(void);int print_cpuinfo (void );#endif int timer_init (void ); -int misc_init_f (void );#if defined(CONFIG_DTB_RESELECT )int embedded_dtb_select (void );
Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
Re: [PATCH v2 09/13] event: Convert misc_init_f() to use events
Quoted 8 lines · Simon Glass
This hook can be implmented using events, for the three boards that actually use it.
Add the event type and event handlers. Drop
CONFIG_MISC_INIT_Fsince we can just useCONFIG_EVENTto control this. Since sandbox always enablesCONFIG_EVENT, we can drop the defconfig lines there too.Signed-off-by Simon Glass <[email protected]>
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
PATCH v2 10/13 event: Convert arch_cpu_init_dm() to use events
In reply to Simon GlassInstead of a special function, send an event after driver model is inited and adjust the boards which use this function.
---
Changes in v2:
- Convert arch/arm/mach-imx/imx8/cpu.c also
- Add DM_EVENT to ARCH_IMX8 too
- Tidy up galileo defconfig and rename quark_init_dm()
- Make a few more of the functions static
- Add a weak function for spl
arch/Kconfig | 3 +++ arch/arm/Kconfig | 4 ++++ arch/arm/mach-imx/imx8/cpu.c | 4 +++- arch/arm/mach-imx/imx8m/soc.c | 4 +++- arch/arm/mach-imx/imx8ulp/soc.c | 4 +++- arch/arm/mach-omap2/am33xx/board.c | 4 +++- arch/arm/mach-omap2/hwinit-common.c | 5 ++++- arch/mips/Kconfig | 1 + arch/mips/mach-pic32/cpu.c | 4 +++- arch/nios2/cpu/cpu.c | 4 +++- arch/riscv/cpu/cpu.c | 5 ++++- arch/riscv/include/asm/system.h | 5 +++++ arch/riscv/lib/spl.c | 3 ++- arch/x86/cpu/baytrail/cpu.c | 4 +++- arch/x86/cpu/broadwell/cpu.c | 4 +++- arch/x86/cpu/ivybridge/cpu.c | 4 +++- arch/x86/cpu/quark/quark.c | 4 +++- arch/x86/include/asm/fsp2/fsp_api.h | 8 ++++++++ arch/x86/lib/fsp2/fsp_init.c | 4 +++- arch/x86/lib/spl.c | 5 +++-- arch/x86/lib/tpl.c | 10 ---------- common/board_f.c | 6 ------ common/event.c | 1 + drivers/core/root.c | 5 +++++ include/event.h | 1 + include/init.h | 11 ----------- 26 files changed, 74 insertions(+), 43 deletions(-)
arch/Kconfig +3 -0
@@ -94,6 +94,7 @@ config NIOS2 bool "Nios II architecture" select CPU select DM + imply DM_EVENT select OF_CONTROL select SUPPORT_OF_CONTROL imply CMD_DM
@@ -113,6 +114,7 @@ config RISCV select DM imply DM_SERIAL imply DM_ETH + imply DM_EVENT imply DM_MMC imply DM_SPI imply DM_SPI_FLASH
@@ -238,6 +240,7 @@ config X86 imply CMD_SF_TEST imply CMD_ZBOOT imply DM_ETH + imply DM_EVENT imply DM_GPIO imply DM_KEYBOARD imply DM_MMC
arch/arm/Kconfig +4 -0
@@ -767,6 +767,7 @@ config ARCH_OMAP2PLUS select SUPPORT_SPL imply TI_SYSC if DM && OF_CONTROL imply FIT + imply DM_EVENT config ARCH_MESON bool "Amlogic Meson"
@@ -811,6 +812,7 @@ config ARCH_IMX8 select MACH_IMX select OF_CONTROL select ENABLE_ARM_SOC_BOOT0_HOOK + imply DM_EVENT config ARCH_IMX8M bool "NXP i.MX8M platform"
@@ -824,6 +826,7 @@ config ARCH_IMX8M select DM select SUPPORT_SPL imply CMD_DM + imply DM_EVENT config ARCH_IMX8ULP bool "NXP i.MX8ULP platform"
@@ -834,6 +837,7 @@ config ARCH_IMX8ULP select SUPPORT_SPL select GPIO_EXTRA_HEADER imply CMD_DM + imply DM_EVENT config ARCH_IMXRT bool "NXP i.MXRT platform"
arch/arm/mach-imx/imx8/cpu.c +3 -1
@@ -8,6 +8,7 @@#include <cpu.h> #include <cpu_func.h> #include <dm.h> +#include <event.h> #include <init.h> #include <log.h> #include <asm/cache.h>
@@ -66,7 +67,7 @@ int arch_cpu_init(void)return 0 ; } -int arch_cpu_init_dm (void ) +static int imx8_init_mu (void *ctx,struct event *event) {struct udevice *devp;int node, ret;
@@ -88,6 +89,7 @@ int arch_cpu_init_dm(void)return 0 ; } +EVENT_SPY (EVT_DM_POST_INIT , imx8_init_mu);int print_bootinfo (void ) {
arch/arm/mach-imx/imx8m/soc.c +3 -1
@@ -7,6 +7,7 @@#include <common.h> #include <cpu_func.h> +#include <event.h> #include <init.h> #include <log.h> #include <asm/arch/imx-regs.h>
@@ -494,7 +495,7 @@ static void imx_set_wdog_powerdown(bool enable)writew (enable, &wdog3->wmcr); } -int arch_cpu_init_dm (void ) +static int imx8m_check_clock (void *ctx,struct event *event) {struct udevice *dev;int ret;
@@ -511,6 +512,7 @@ int arch_cpu_init_dm(void)return 0 ; } +EVENT_SPY (EVT_DM_POST_INIT , imx8m_check_clock);int arch_cpu_init (void ) {
arch/arm/mach-imx/imx8ulp/soc.c +3 -1
@@ -11,6 +11,7 @@#include <asm/mach-imx/boot_mode.h> #include <asm/global_data.h> #include <efi_loader.h> +#include <event.h> #include <spl.h> #include <asm/arch/rdc.h> #include <asm/arch/s400_api.h>
@@ -569,7 +570,7 @@ int arch_cpu_init(void)return 0 ; } -int arch_cpu_init_dm (void ) +static int imx8ulp_check_mu (void *ctx,struct event *event) {struct udevice *devp;int node, ret;
@@ -584,6 +585,7 @@ int arch_cpu_init_dm(void)return 0 ; } +EVENT_SPY (EVT_DM_POST_INIT , imx8ulp_check_mu);#if defined(CONFIG_SPL_BUILD )__weak void __noreturn jump_to_image_no_args (struct spl_image_info *spl_image)
arch/arm/mach-omap2/am33xx/board.c +3 -1
@@ -11,6 +11,7 @@#include <dm.h> #include <debug_uart.h> #include <errno.h> +#include <event.h> #include <init.h> #include <net.h> #include <ns16550.h>
@@ -596,7 +597,7 @@ void board_init_f(ulong dummy)#endif -int arch_cpu_init_dm (void ) +static int am33xx_dm_post_init (void *ctx,struct event *event) {hw_data_init ();#if !CONFIG_IS_ENABLED (SKIP_LOWLEVEL_INIT )
@@ -604,3 +605,4 @@ int arch_cpu_init_dm(void)#endif return 0 ; } +EVENT_SPY (EVT_DM_POST_INIT , am33xx_dm_post_init);
arch/arm/mach-omap2/hwinit-common.c +4 -1
@@ -12,6 +12,7 @@*/ #include <common.h> #include <debug_uart.h> +#include <event.h> #include <fdtdec.h> #include <init.h> #include <spl.h>
@@ -239,11 +240,13 @@ void board_init_f(ulong dummy) }#endif -int arch_cpu_init_dm (void ) +static int omap2_system_init (void *ctx,struct event *event) {early_system_init (); +return 0 ; } +EVENT_SPY (EVT_DM_POST_INIT , omap2_system_init);/* * Routine: wait_for_command_complete
arch/mips/Kconfig +1 -0
@@ -130,6 +130,7 @@ config MACH_PIC32 config TARGET_BOSTON bool "Support Boston" select DM + imply DM_EVENT select DM_SERIAL select MIPS_CM select SYS_CACHE_SHIFT_6
arch/mips/mach-pic32/cpu.c +3 -1
@@ -7,6 +7,7 @@#include <common.h> #include <clk.h> #include <dm.h> +#include <event.h> #include <init.h> #include <malloc.h> #include <asm/global_data.h>
@@ -95,12 +96,13 @@ static void prefetch_init(void) }/* arch specific CPU init after DM */ -int arch_cpu_init_dm (void ) +static int pic32_flash_prefetch (void *ctx,struct event *event) {/* flash prefetch */ prefetch_init ();return 0 ; } +EVENT_SPY (EVT_DM_POST_INIT , pic32_flash_prefetch);/* Un-gate DDR2 modules (gated by default) */ static void ddr2_pmd_ungate (void )
arch/nios2/cpu/cpu.c +3 -1
@@ -10,6 +10,7 @@#include <cpu_func.h> #include <dm.h> #include <errno.h> +#include <event.h> #include <init.h> #include <irq_func.h> #include <asm/cache.h>
@@ -63,7 +64,7 @@ static void copy_exception_trampoline(void) }#endif -int arch_cpu_init_dm (void ) +static int nios_cpu_setup (void *ctx,struct event *event) {struct udevice *dev;int ret;
@@ -79,6 +80,7 @@ int arch_cpu_init_dm(void)return 0 ; } +EVENT_SPY (EVT_DM_POST_INIT , nios_cpu_setup);static int altera_nios2_get_desc (const struct udevice *dev,char *buf,int size)
arch/riscv/cpu/cpu.c +4 -1
@@ -7,9 +7,11 @@#include <cpu.h> #include <dm.h> #include <dm/lists.h> +#include <event.h> #include <init.h> #include <log.h> #include <asm/encoding.h> +#include <asm/system.h> #include <dm/uclass-internal.h> #include <linux/bitops.h>
@@ -81,7 +83,7 @@ static void dummy_pending_ipi_clear(ulong hart, ulong arg0, ulong arg1) }#endif -int arch_cpu_init_dm (void ) +int riscv_cpu_setup (void *ctx,struct event *event) {int ret;
@@ -133,6 +135,7 @@ int arch_cpu_init_dm(void)return 0 ; } +EVENT_SPY (EVT_DM_POST_INIT , riscv_cpu_setup);int arch_early_init_r (void ) {
arch/riscv/include/asm/system.h +5 -0
@@ -7,6 +7,8 @@#ifndef __ASM_RISCV_SYSTEM_H#define __ASM_RISCV_SYSTEM_H +struct event ; +/* * Interrupt configuring macros. *
@@ -14,4 +16,7 @@* */ +/* Hook to set up the CPU (called from SPL too) */ +int riscv_cpu_setup (void *ctx,struct event *event); +#endif /* __ASM_RISCV_SYSTEM_H */
arch/riscv/lib/spl.c +2 -1
@@ -11,6 +11,7 @@#include <spl.h> #include <asm/global_data.h> #include <asm/smp.h> +#include <asm/system.h> DECLARE_GLOBAL_DATA_PTR ;
@@ -27,7 +28,7 @@ __weak void board_init_f(ulong dummy)if (ret)panic ("spl_early_init() failed: %d\n" , ret); -arch_cpu_init_dm (); +riscv_cpu_setup (NULL ,NULL );preloader_console_init ();
arch/x86/cpu/baytrail/cpu.c +3 -1
@@ -8,6 +8,7 @@#include <common.h> #include <cpu.h> #include <dm.h> +#include <event.h> #include <init.h> #include <log.h> #include <pci.h>
@@ -44,7 +45,7 @@ static void hsuart_clock_set(void *base)* Configure the internal clock of both SIO HS-UARTs, if they are enabled * via FSP */ -int arch_cpu_init_dm (void ) +static int baytrail_uart_init (void *ctx,struct event *event) {struct udevice *dev;void *base;
@@ -63,6 +64,7 @@ int arch_cpu_init_dm(void)return 0 ; } +EVENT_SPY (EVT_DM_POST_INIT , baytrail_uart_init);static void set_max_freq (void ) {
arch/x86/cpu/broadwell/cpu.c +3 -1
@@ -8,6 +8,7 @@#include <common.h> #include <dm.h> #include <cpu.h> +#include <event.h> #include <init.h> #include <log.h> #include <asm/cpu.h>
@@ -24,7 +25,7 @@#include <asm/arch/pch.h> #include <asm/arch/rcb.h> -int arch_cpu_init_dm (void ) +static int broadwell_init_cpu (void *ctx,struct event *event) {struct udevice *dev;int ret;
@@ -41,6 +42,7 @@ int arch_cpu_init_dm(void)return 0 ; } +EVENT_SPY (EVT_DM_POST_INIT , broadwell_init_cpu);void set_max_freq (void ) {
arch/x86/cpu/ivybridge/cpu.c +3 -1
@@ -14,6 +14,7 @@#include <cpu_func.h> #include <dm.h> #include <errno.h> +#include <event.h> #include <fdtdec.h> #include <init.h> #include <log.h>
@@ -53,7 +54,7 @@ int arch_cpu_init(void)return x86_cpu_init_f (); } -int arch_cpu_init_dm (void ) +static int ivybridge_cpu_init (void *ctx,struct event *ev) {struct pci_controller *hose;struct udevice *bus, *dev;
@@ -85,6 +86,7 @@ int arch_cpu_init_dm(void)return 0 ; } +EVENT_SPY (EVT_DM_POST_INIT , ivybridge_cpu_init);#define PCH_EHCI0_TEMP_BAR0 0xe8000000#define PCH_EHCI1_TEMP_BAR0 0xe8000400
arch/x86/cpu/quark/quark.c +3 -1
@@ -5,6 +5,7 @@#include <common.h> #include <cpu_func.h> +#include <event.h> #include <init.h> #include <mmc.h> #include <asm/cache.h>
@@ -247,7 +248,7 @@ int arch_cpu_init(void)return 0 ; } -int arch_cpu_init_dm (void ) +static int quark_init_pcie (void *ctx,struct event *event) {/* * Initialize PCIe controller
@@ -262,6 +263,7 @@ int arch_cpu_init_dm(void)return 0 ; } +EVENT_SPY (EVT_DM_POST_INIT , quark_init_pcie);int checkcpu (void ) {
arch/x86/include/asm/fsp2/fsp_api.h +8 -0
@@ -60,4 +60,12 @@ int fsp_silicon_init(bool s3wake, bool use_spi_flash); typedef asmlinkage int (*fsp_silicon_init_func)(struct fsps_upd *params); +/** +* fsp_setup_pinctrl() - Set up the pinctrl for FSP +* +* @ctx: Event context (not used) +* @event: Event information (not used) +*/ +int fsp_setup_pinctrl (void *ctx,struct event *event); +#endif
arch/x86/lib/fsp2/fsp_init.c +3 -1
@@ -9,6 +9,7 @@#include <bootstage.h> #include <cbfs.h> #include <dm.h> +#include <event.h> #include <init.h> #include <log.h> #include <spi.h>
@@ -18,7 +19,7 @@#include <dm/uclass-internal.h> #include <asm/fsp2/fsp_internal.h> -int arch_cpu_init_dm (void ) +int fsp_setup_pinctrl (void *ctx,struct event *event) {struct udevice *dev;ofnode node;
@@ -41,6 +42,7 @@ int arch_cpu_init_dm(void)return ret; } +EVENT_SPY (EVT_DM_POST_INIT , fsp_setup_pinctrl);#if !defined(CONFIG_TPL_BUILD )binman_sym_declare (ulong, intel_fsp_m, image_pos);
arch/x86/lib/spl.c +3 -2
@@ -17,6 +17,7 @@#include <syscon.h> #include <asm/cpu.h> #include <asm/cpu_common.h> +#include <asm/fsp2/fsp_api.h> #include <asm/global_data.h> #include <asm/mrccache.h> #include <asm/mtrr.h>
@@ -27,7 +28,7 @@DECLARE_GLOBAL_DATA_PTR ; -__weak int arch_cpu_init_dm (void ) +__weak int fsp_setup_pinctrl (void *ctx,struct event *event) {return 0 ; }
@@ -89,7 +90,7 @@ static int x86_spl_init(void)return ret; }#ifndef CONFIG_TPL - ret =arch_cpu_init_dm (); + ret =fsp_setup_pinctrl (NULL ,NULL );if (ret) {debug ("%s: arch_cpu_init_dm() failed\n" , __func__);return ret;
arch/x86/lib/tpl.c +0 -10
@@ -19,11 +19,6 @@DECLARE_GLOBAL_DATA_PTR ; -__weak int arch_cpu_init_dm (void ) -{ -return 0 ; -} -static int x86_tpl_init (void ) {int ret;
@@ -44,11 +39,6 @@ static int x86_tpl_init(void)debug ("%s: arch_cpu_init() failed\n" , __func__);return ret; } - ret =arch_cpu_init_dm (); -if (ret) { -debug ("%s: arch_cpu_init_dm() failed\n" , __func__); -return ret; - }preloader_console_init ();return 0 ;
common/board_f.c +0 -6
@@ -803,11 +803,6 @@ __weak int reserve_arch(void)return 0 ; } -__weak int arch_cpu_init_dm (void ) -{ -return 0 ; -} -__weak int checkcpu (void ) {return 0 ;
@@ -848,7 +843,6 @@ static const init_fnc_t init_sequence_f[] = { arch_cpu_init,/* basic arch cpu dependent setup */ mach_cpu_init,/* SoC/machine dependent CPU setup */ initf_dm, - arch_cpu_init_dm, #if defined(CONFIG_BOARD_EARLY_INIT_F) board_early_init_f, #endif
common/event.c +1 -0
@@ -26,6 +26,7 @@ const char *const type_name[] = {"test" ,/* Events related to driver model */ +"dm_post_init" ,"dm_pre_probe" ,"dm_post_probe" , "dm_pre_remove",
drivers/core/root.c +5 -0
@@ -404,6 +404,11 @@ int dm_init_and_scan(bool pre_reloc_only)return ret; } } +if (CONFIG_IS_ENABLED (DM_EVENT )) { + ret =event_notify_null (EVT_DM_POST_INIT ); +if (ret) +return log_msg_ret ("ev" , ret); + }return 0 ; }
include/event.h +1 -0
@@ -20,6 +20,7 @@ enum event_t {EVT_TEST ,/* Events related to driver model */ +EVT_DM_POST_INIT ,EVT_DM_PRE_PROBE ,EVT_DM_POST_PROBE , EVT_DM_PRE_REMOVE,
include/init.h +0 -11
@@ -45,17 +45,6 @@ void board_init_f(ulong dummy);*/ int arch_cpu_init (void ); -/** -* arch_cpu_init_dm() - init CPU after driver model is available -* -* This is called immediately after driver model is available before -* relocation. This is similar to arch_cpu_init() but is able to reference -* devices -* -* Return: 0 if OK, -ve on error -*/ -int arch_cpu_init_dm (void ); -/** * mach_cpu_init() - SoC/machine dependent CPU setup *
Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
Re: [PATCH v2 10/13] event: Convert arch_cpu_init_dm() to use events
Quoted 4 lines · Simon Glass
Instead of a special function, send an event after driver model is inited and adjust the boards which use this function.
Signed-off-by Simon Glass <[email protected]>
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
PATCH v2 11/13 event: Add a command
In reply to Simon GlassAdd a command to show the available events.
---
(no changes since v1)
MAINTAINERS | 1 + cmd/Kconfig | 8 ++++++++ cmd/Makefile | 1 + cmd/event.c | 27 +++++++++++++++++++++++++++ include/event.h | 3 +++ 5 files changed, 40 insertions(+) create mode 100644 cmd/event.c
MAINTAINERS +1 -0
@@ -812,6 +812,7 @@ F: scripts/env2string.awk EVENTS M: Simon Glass <[email protected]> S: Maintained +F: cmd/event.c F: common/event.c F: include/event.h F: test/common/event.c
cmd/Kconfig +8 -0
@@ -2342,6 +2342,14 @@ config CMD_DIAG available tests and running either all the tests, or specific tests identified by name. +config CMD_EVENT + bool "event - Show information about events" + default y if EVENT_DEBUG + help + This enables the 'event' command which provides information about + events and event-handler routines. This can help to device event + hadling. + config CMD_IRQ bool "irq - Show information about interrupts" depends on !ARM && !MIPS && !RISCV && !SH
cmd/Makefile +1 -0
@@ -53,6 +53,7 @@ obj-$(CONFIG_CMD_DIAG) += diag.o endif obj-$(CONFIG_CMD_ADTIMG) += adtimg.o obj-$(CONFIG_CMD_ABOOTIMG) += abootimg.o +obj-$(CONFIG_CMD_EVENT) += event.o obj-$(CONFIG_CMD_EXTENSION) += extension_board.o obj-$(CONFIG_CMD_ECHO) += echo.o obj-$(CONFIG_ENV_IS_IN_EEPROM) += eeprom.o
cmd/event.c new file +27 -0
@@ -0,0 +1,27 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* Command-line access to events +* +* Copyright 2021 Google LLC +* Written by Simon Glass <[email protected]> +*/ + +#include <common.h> +#include <command.h> +#include <event.h> + +static int do_event_list (struct cmd_tbl *cmdtp,int flag,int argc, +char *const argv[]) +{ +event_show_spy_list (); + +return 0 ; +} + +#ifdef CONFIG_SYS_LONGHELP +static char event_help_text[] = +"event list - list event spies" ; +#endif + +U_BOOT_CMD_WITH_SUBCMDS (event,"Events" , event_help_text, +U_BOOT_SUBCMD_MKENT (list,1 ,1 , do_event_list));
include/event.h +3 -0
@@ -141,6 +141,9 @@ static inline const char *event_spy_id(struct evspy_info *spy)int event_register (const char *id,enum event_t type,event_handler_t func,void *ctx); +/** event_show_spy_list( - Show a list of event spies */ +void event_show_spy_list (void ); +#if CONFIG_IS_ENABLED (EVENT )/** * event_notify() - notify spies about an event
Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
Re: [PATCH v2 11/13] event: Add a command
Quoted 3 lines · Simon Glass
Add a command to show the available events.
Signed-off-by Simon Glass <[email protected]>
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
PATCH v2 12/13 event: Add a script to decode the event-spy list
In reply to Simon GlassFor debugging and dicoverability it is useful to be able to see a list of each event spy in a U-Boot ELF file. Add a script which shows this, along with the event type and the source location. This makes events a little easier to use than weak functions, for example.
Add a basic sandbox test as well. We could provide a test for other boards, but for now, few use events.
---
Changes in v2:
- Update for patman snake-case change
- Use a regular expression in the test to avoid dependency on build dir
MAINTAINERS | 2 + scripts/event_dump.py | 115 +++++++++++++++++++++++++++++++ test/py/tests/test_event_dump.py | 20 ++++++ 3 files changed, 137 insertions(+) create mode 100755 scripts/event_dump.py create mode 100644 test/py/tests/test_event_dump.py
MAINTAINERS +2 -0
@@ -815,7 +815,9 @@ S: Maintained F: cmd/event.c F: common/event.c F: include/event.h +F: scripts/event_dump.py F: test/common/event.c +F: test/py/tests/test_event_dump.py FASTBOOT S: Orphaned
scripts/event_dump.py new file +115 -0
115 lines not on this page · raw message
test/py/tests/test_event_dump.py new file +20 -0
@@ -0,0 +1,20 @@ +# SPDX-License-Identifier: GPL-2.0+ +# Copyright 2021 Google LLC +# Written by Simon Glass <[email protected]> + +import pytest +import re +import u_boot_utils as util + +# This is only a partial test - coverting 64-bit sandbox. It does not test +# big-endian images, nor 32-bit images [email protected]('sandbox') +def test_event_dump(u_boot_console): + """Test that the "help" command can be executed.""" + cons = u_boot_console + sandbox = cons.config.build_dir + '/u-boot' + out = util.run_and_log(cons, ['scripts/event_dump.py', sandbox]) + expect = '''.*Event type Id Source location +-------------------- ------------------------------ ------------------------------ +EVT_MISC_INIT_F sandbox_misc_init_f .*arch/sandbox/cpu/start.c:''' + assert re.match(expect, out, re.MULTILINE) is not None
Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
Re: [PATCH v2 12/13] event: Add a script to decode the event-spy list
Quoted 9 lines · Simon Glass
For debugging and dicoverability it is useful to be able to see a list of each event spy in a U-Boot ELF file. Add a script which shows this, along with the event type and the source location. This makes events a little easier to use than weak functions, for example.
Add a basic sandbox test as well. We could provide a test for other boards, but for now, few use events.
Signed-off-by Simon Glass <[email protected]>
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
PATCH v2 13/13 event: Add documentation
In reply to Simon GlassAdd documentation for events, including the event command.
---
(no changes since v1)
doc/develop/event.rst | 66 +++++++++++++++++++++++++++++++++++++++++++ doc/develop/index.rst | 1 + doc/usage/event.rst | 49 ++++++++++++++++++++++++++++++++ doc/usage/index.rst | 1 + 4 files changed, 117 insertions(+) create mode 100644 doc/develop/event.rst create mode 100644 doc/usage/event.rst
doc/develop/event.rst new file +66 -0
@@ -0,0 +1,66 @@ +.. SPDX-License-Identifier: GPL-2.0+ + +Events +====== + +U-Boot supports a way for various events to be handled by interested +subsystems. This provide a generic way to handle 'hooks' like setting up the +CPUs after driver model is active, or reading a partition table after a new +block device is probed. + +Rather than using weak functions and direct calls across subsystemss, it is +often easier to use an event. + +An event consists of a type (e.g. EVT_DM_POST_INIT) and some optional data, +in `union event_data`. An event spy can be creasted to watch for events of a +particular type. When the event is created, it is sent to each spy in turn. + + +Declaring a spy +--------------- + +To declare a spy, use something like this:: + + static int snow_setup_cpus(void *ctx, struct event *event) + { + /* do something */ + return 0; + } + EVENT_SPY(EVT_DM_POST_INIT, snow_setup_cpus); + +Your function is called when EVT_DM_POST_INIT is emitted, i.e. after driver +model is inited (in SPL, or in U-Boot proper before and after relocation). + + +Debugging +--------- + +To assist with debugging events, enable `CONFIG_EVENT_DEBUG` and +`CONFIG_CMD_EVENT`. The :doc:`../usage/event` command can then be used to +provide a spy list. + +It is also possible to list spy information from the U-Boot executable,, using +the `event_dump.py` script:: + + $ scripts/event_dump.py /tmp/b/sandbox/u-boot + Event type Id Source location + -------------------- ------------------------------ ------------------------------ + EVT_MISC_INIT_F f:sandbox_misc_init_f arch/sandbox/cpu/start.c:125 + +This shows each event spy in U-Boot, along with the event type, function name +(or ID) and source location. + +Note that if `CONFIG_EVENT_DEBUG` is not enabled, the event ID is missing, so +the function is shown instead (with an `f:` prefix as above). Since the ID is +generally the same as the function name, this does not matter much. + +The event type is decoded by the symbol used by U-Boot for the event linker +list. Symbols have the form:: + + _u_boot_list_2_evspy_info_2_EVT_MISC_INIT_F + +so the event type can be read from the end. To manually list spy information +in an image, use $(CROSS_COMPILE)nm:: + + nm u-boot |grep evspy |grep list + 00000000002d6300 D _u_boot_list_2_evspy_info_2_EVT_MISC_INIT_F
doc/develop/index.rst +1 -0
@@ -17,6 +17,7 @@ Implementation distro driver-model/index environment + event global_data logging makefiles
doc/usage/event.rst new file +49 -0
49 lines not on this page · raw message
doc/usage/index.rst +1 -0
@@ -29,6 +29,7 @@ Shell commands x86/cbsysinfo conitrace echo + event exception extension exit
Signature · 1 line
2.35.1.616.g0bdcbb4464-goog
Re: [PATCH v2 13/13] event: Add documentation
Quoted 3 lines · Simon Glass
Add documentation for events, including the event command.
Signed-off-by Simon Glass <[email protected]>
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
Quoted 15 lines · Simon Glass
It is a common need in U-Boot to have one subsystem notify another when something happens. An example is reading a partition table when a new block device is set up.
It is also common to add weak functions and 'hook' functions to modify how U-Boot works. See for example
ft_board_setup()and the like.U-Boot would benefit from a generic mechanism to handle these cases, with the ability to hook into various 'events' in a subsystem-independent and transparent way.
This series provides a way to create and dispatch events, with a way of registering a 'spy' which watches for events of different types. This allows 'hook' functions to be created in a generic way.
Quoted 8 lines
It also includes a script to list the hooks in an image, which is a bit easier to debug than weak functions, as well as an 'event' command to do the same from within U-Boot.
These 'static' events can be used to replace hooks like
misc_init_f(), for example. Also included is basic support for 'dynamic' events, where a spy can be registered at runtime. The need for this is still being figured out.
@Simon, Tom:
What is the status of this series? Takahiro's UEFI integration series builds on it. Is it going to be pushed to origin/next soon?
Best regards
Heinrich
Quoted 113 lines
Changes in v2:
- Add a 'used' attribute to avoid LTO dropping the code
- Update keymile pg-wcom boards also
- Make the sandbox function static
- Convertarch/arm/mach-imx/imx8/cpu.calso
- AddDM_EVENTtoARCH_IMX8too
- Tidy up galileo defconfig and renamequark_init_dm()
- Make a few more of the functions static
- Add a weak function for spl
- Update for patman snake-case change
- Use a regular expression in the test to avoid dependency on build dirSimon Glass (12):
Makefile: Allow LTO to be disabled for a build
sandbox: start: Sort the header files
binman: Expand elf support a little
event: Add basic support for events
event: Add a simple test
event: Set up the event system on start-up
event: Add events for device probe/remove
event: Convertmisc_init_f()to use events
event: Convertarch_cpu_init_dm()to use events
event: Add a command
event: Add a script to decode the event-spy list
event: Add documentationTim Harvey (1):
phy: nop-phy: Fix phy reset if no reset-gpio definedMAINTAINERS | 10 + Makefile | 18 +- arch/Kconfig | 3 + arch/arm/Kconfig | 4 + arch/arm/config.mk | 4 +- arch/arm/include/asm/global_data.h | 2 +- arch/arm/mach-imx/imx8/cpu.c | 4 +- arch/arm/mach-imx/imx8m/soc.c | 4 +- arch/arm/mach-imx/imx8ulp/soc.c | 4 +- arch/arm/mach-omap2/am33xx/board.c | 4 +- arch/arm/mach-omap2/hwinit-common.c | 5 +- arch/mips/Kconfig | 1 + arch/mips/mach-pic32/cpu.c | 4 +- arch/nios2/cpu/cpu.c | 4 +- arch/riscv/cpu/cpu.c | 5 +- arch/riscv/include/asm/system.h | 5 + arch/riscv/lib/spl.c | 3 +- arch/sandbox/cpu/start.c | 8 +- arch/x86/cpu/baytrail/cpu.c | 4 +- arch/x86/cpu/broadwell/cpu.c | 4 +- arch/x86/cpu/ivybridge/cpu.c | 4 +- arch/x86/cpu/quark/quark.c | 4 +- arch/x86/include/asm/fsp2/fsp_api.h | 8 + arch/x86/lib/fsp2/fsp_init.c | 4 +- arch/x86/lib/spl.c | 5 +- arch/x86/lib/tpl.c | 10 - board/google/chromebook_coral/coral.c | 7 +- board/keymile/kmcent2/kmcent2.c | 4 +- .../keymile/pg-wcom-ls102xa/pg-wcom-ls102xa.c | 5 +- cmd/Kconfig | 8 + cmd/Makefile | 1 + cmd/event.c | 27 +++ common/Kconfig | 37 ++- common/Makefile | 2 + common/board_f.c | 13 +- common/board_r.c | 1 + common/event.c | 196 ++++++++++++++++ common/log.c | 1 + configs/chromebook_coral_defconfig | 1 + configs/kmcent2_defconfig | 2 +- configs/pg_wcom_expu1_defconfig | 1 + configs/pg_wcom_expu1_update_defconfig | 1 + configs/pg_wcom_seli8_defconfig | 1 + configs/pg_wcom_seli8_update_defconfig | 1 + configs/sandbox64_defconfig | 1 - configs/sandbox_defconfig | 1 - configs/sandbox_flattree_defconfig | 1 - configs/sandbox_spl_defconfig | 1 - configs/tools-only_defconfig | 1 - doc/build/gcc.rst | 17 ++ doc/develop/event.rst | 66 ++++++ doc/develop/index.rst | 1 + doc/usage/event.rst | 49 ++++ doc/usage/index.rst | 1 + drivers/core/Kconfig | 10 + drivers/core/device-remove.c | 8 + drivers/core/device.c | 9 + drivers/core/root.c | 5 + drivers/phy/nop-phy.c | 12 +- include/asm-generic/global_data.h | 13 ++ include/configs/km/pg-wcom-ls102xa.h | 2 - include/dm/device-internal.h | 10 + include/event.h | 210 ++++++++++++++++++ include/event_internal.h | 35 +++ include/init.h | 12 - include/log.h | 2 + scripts/event_dump.py | 115 ++++++++++ test/common/Makefile | 1 + test/common/event.c | 85 +++++++ test/py/tests/test_event_dump.py | 20 ++ test/test-main.c | 4 + tools/binman/elf.py | 58 ++++- 72 files changed, 1108 insertions(+), 86 deletions(-)create mode 100644
cmd/event.c
create mode 100644common/event.c
create mode 100644doc/develop/event.rstcreate mode 100644doc/usage/event.rstcreate mode 100644include/event.h
create mode 100644include/event_internal.hcreate mode 100755scripts/event_dump.pycreate mode 100644test/common/event.ccreate mode 100644test/py/tests/test_event_dump.py
Quoted 25 lines · Heinrich Schuchardt
Quoted 15 lines · Simon Glass
It is a common need in U-Boot to have one subsystem notify another when something happens. An example is reading a partition table when a new block device is set up.
It is also common to add weak functions and 'hook' functions to modify how U-Boot works. See for example
ft_board_setup()and the like.U-Boot would benefit from a generic mechanism to handle these cases, with the ability to hook into various 'events' in a subsystem-independent and transparent way.
This series provides a way to create and dispatch events, with a way of registering a 'spy' which watches for events of different types. This allows 'hook' functions to be created in a generic way.
Quoted 8 lines
It also includes a script to list the hooks in an image, which is a bit easier to debug than weak functions, as well as an 'event' command to do the same from within U-Boot.
These 'static' events can be used to replace hooks like
misc_init_f(), for example. Also included is basic support for 'dynamic' events, where a spy can be registered at runtime. The need for this is still being figured out.
Quoted 4 lines
@Simon, Tom:
What is the status of this series? Takahiro's UEFI integration series builds on it. Is it going to be pushed to origin/next soon?
I'm waiting for Simon to reply to Takahiro's comment on v2 before applying.
Signature · 1 line
Tom
Hi Tom,
Quoted 27 lines · Tom Rini
Quoted 25 lines · Heinrich Schuchardt
Quoted 23 lines · Simon Glass
It is a common need in U-Boot to have one subsystem notify another when something happens. An example is reading a partition table when a new block device is set up.
It is also common to add weak functions and 'hook' functions to modify how U-Boot works. See for example
ft_board_setup()and the like.U-Boot would benefit from a generic mechanism to handle these cases, with the ability to hook into various 'events' in a subsystem-independent and transparent way.
This series provides a way to create and dispatch events, with a way of registering a 'spy' which watches for events of different types. This allows 'hook' functions to be created in a generic way.
It also includes a script to list the hooks in an image, which is a bit easier to debug than weak functions, as well as an 'event' command to do the same from within U-Boot.
These 'static' events can be used to replace hooks like
misc_init_f(), for example. Also included is basic support for 'dynamic' events, where a spy can be registered at runtime. The need for this is still being figured out.
Quoted 7 lines
Quoted 4 lines
@Simon, Tom:
What is the status of this series? Takahiro's UEFI integration series builds on it. Is it going to be pushed to origin/next soon?
I'm waiting for Simon to reply to Takahiro's comment on v2 before applying.
OK I found it and replied.
Regards,
Simon