[PATCH v4] firmware: remove FW_TEXT_START
Jessica Clarke
jrtc27 at jrtc27.com
Tue May 14 09:56:45 PDT 2024
On 14 May 2024, at 17:33, Anup Patel <anup at brainfault.org> wrote:
>
> On Tue, May 14, 2024 at 9:58 PM Clément Léger <cleger at rivosinc.com> wrote:
>>
>>
>>
>> On 10/04/2024 07:18, Anup Patel wrote:
>>> On Mon, Apr 8, 2024 at 9:01 PM Xiang W <wxjstz at 126.com> wrote:
>>>>
>>>> Now opensbi can run at any address via dynamic relocation. We can
>>>> remove FW_TEXT_START.
>>>>
>>>> Signed-off-by: Xiang W <wxjstz at 126.com>
>>>
>>> LGTM.
>>>
>>> Reviewed-by: Anup Patel <anup at brainfault.org>
>>> Tested-by: Anup Patel <anup at brainfault.org>
>>>
>>> Applied this patch to the riscv/opensbi repo.
>>>
>>> Thanks,
>>> Anup
>>
>> Hi Anup,
>>
>> This patch seems to break spike support. The newly created ELF is not
>> marked as EXEC anymore but DYNAMIC and it fails with the following error:
>>
>> spike: ../fesvr/elfloader.cc:45:
>> std::map<std::__cxx11::basic_string<char>, long unsigned int>
>> load_elf(const char*, memif_t*, reg_t*, unsigned int): Assertion
>> `IS_ELF_EXEC(*eh64)' failed.
>>
>> Should we fixed OpenSBI or allow spike to load a DYNAMIC elf ?
>
> I think it is better to use FW_DYNAMIC in Spike because various
> other projects (such as U-Boot, QEMU, etc) use it.
That’s an orthogonal concern. This is about the ELF file type, not the
boot method used, but supporting more boot methods is good too.
Spike should support ET_DYN files, but will need to adjust all the load
addresses as they’ll be zero, so it’s not just a case of adding an
option to the assert.
> Also, Spike should support loading M-mode firmware in BIN
> (flat binary) format as well.
There’s not really a good reason to prefer a flat binary over an ELF
for an emulator*. For a board it makes sense because you want to have
the flat binary ready to go in memory, but for an emulator it’s more
useful to have the richer structure that an ELF provides, especially if
that includes debug info. Historically Spike needed it to be an ELF so
it could extract the HTIF symbol information as otherwise you’d have no
console I/O, but I believe these days Spike can emulate an 8250 so that
may no longer be a requirement?
Jess
* Space savings are minimal to non-existent; Debian’s U-Boot
qemu-riscv64_smode saves 0.8%, and the generic OpenSBI flat
binaries are bigger than the ELF files
> Regards,
> Anup
>
>>
>> Thanks,
>>
>> Clément
>>
>>>
>>>> ---
>>>>
>>>> v4 changes:
>>>> - only remove FW_TEXT_START by anup's suggestion
>>>>
>>>> docs/firmware/fw.md | 2 --
>>>> docs/firmware/fw_jump.md | 4 ++--
>>>> docs/firmware/fw_payload.md | 6 +++---
>>>> docs/platform/generic.md | 11 +++--------
>>>> firmware/fw_base.S | 4 +---
>>>> firmware/fw_base.ldS | 3 +--
>>>> firmware/fw_payload.elf.ldS | 2 +-
>>>> firmware/objects.mk | 4 ----
>>>> firmware/payloads/test.elf.ldS | 2 +-
>>>> platform/fpga/ariane/objects.mk | 1 -
>>>> platform/fpga/openpiton/objects.mk | 1 -
>>>> platform/generic/objects.mk | 3 +--
>>>> platform/kendryte/k210/objects.mk | 1 -
>>>> platform/nuclei/ux600/objects.mk | 1 -
>>>> platform/template/objects.mk | 9 ++-------
>>>> 15 files changed, 15 insertions(+), 39 deletions(-)
>>>>
>>>> diff --git a/docs/firmware/fw.md b/docs/firmware/fw.md
>>>> index 2f4deb5..3cc0262 100644
>>>> --- a/docs/firmware/fw.md
>>>> +++ b/docs/firmware/fw.md
>>>> @@ -61,8 +61,6 @@ Firmware Configuration and Compilation
>>>> All firmware types support the following common compile time configuration
>>>> parameters:
>>>>
>>>> -* **FW_TEXT_START** - Defines the execution address of the OpenSBI firmware.
>>>> - This configuration parameter is mandatory.
>>>> * **FW_FDT_PATH** - Path to an external flattened device tree binary file to
>>>> be embedded in the *.rodata* section of the final firmware. If this option
>>>> is not provided then the firmware will expect the FDT to be passed as an
>>>> diff --git a/docs/firmware/fw_jump.md b/docs/firmware/fw_jump.md
>>>> index 2ee6b29..66be016 100644
>>>> --- a/docs/firmware/fw_jump.md
>>>> +++ b/docs/firmware/fw_jump.md
>>>> @@ -35,7 +35,7 @@ follows:
>>>> At least one of *FW_JUMP_ADDR* and *FW_JUMP_OFFSET* (see below) should be
>>>> defined. Compilation errors will result from not defining one of them.
>>>>
>>>> -* **FW_JUMP_OFFSET** - Address offset from the *FW_TEXT_START* where the
>>>> +* **FW_JUMP_OFFSET** - Address offset from the opensbi load address where the
>>>> entry point of the next booting stage is located. This offset is used as
>>>> relocatable address of the next booting stage entry point. If *FW_JUMP_ADDR*
>>>> is also defined, the firmware will prefer *FW_JUMP_ADDR*.
>>>> @@ -62,7 +62,7 @@ follows:
>>>> echo fdt overlaps kernel, increase FW_JUMP_FDT_ADDR
>>>> ```
>>>>
>>>> -* **FW_JUMP_FDT_OFFSET** - Address offset from the *FW_TEXT_START* where
>>>> +* **FW_JUMP_FDT_OFFSET** - Address offset from the opensbi load address where
>>>> the FDT will be passed to the next booting stage. This offset is used
>>>> as relocatable address of the FDT passed to the next booting stage. If
>>>> *FW_JUMP_FDT_ADDR* is also defined, the firmware will prefer
>>>> diff --git a/docs/firmware/fw_payload.md b/docs/firmware/fw_payload.md
>>>> index a67fc50..d25be60 100644
>>>> --- a/docs/firmware/fw_payload.md
>>>> +++ b/docs/firmware/fw_payload.md
>>>> @@ -36,8 +36,8 @@ options. These configuration parameters can be defined using either the top
>>>> level `make` command line or the target platform *objects.mk* configuration
>>>> file. The parameters currently defined are as follows:
>>>>
>>>> -* **FW_PAYLOAD_OFFSET** - Offset from *FW_TEXT_START* where the payload binary
>>>> - will be linked in the final *FW_PAYLOAD* firmware binary image. This
>>>> +* **FW_PAYLOAD_OFFSET** - Offset from the opensbi load address where the payload
>>>> + binary will be linked in the final *FW_PAYLOAD* firmware binary image. This
>>>> configuration parameter is mandatory if *FW_PAYLOAD_ALIGN* is not defined.
>>>> Compilation errors will result from an incorrect definition of
>>>> *FW_PAYLOAD_OFFSET* or of *FW_PAYLOAD_ALIGN*, or if neither of these
>>>> @@ -62,7 +62,7 @@ file. The parameters currently defined are as follows:
>>>> firmware will pass the FDT address passed by the previous booting stage
>>>> to the next booting stage.
>>>>
>>>> -* **FW_PAYLOAD_FDT_OFFSET** - Address offset from the *FW_TEXT_START* where
>>>> +* **FW_PAYLOAD_FDT_OFFSET** - Address offset from the opensbi load address where
>>>> the FDT will be passed to the next booting stage. This offset is used as
>>>> relocatable address of the FDT passed to the next booting stage. If
>>>> *FW_PAYLOAD_FDT_ADDR* is also defined, the firmware will prefer *FW_PAYLOAD_FDT_ADDR*.
>>>> diff --git a/docs/platform/generic.md b/docs/platform/generic.md
>>>> index c29eb04..709b436 100644
>>>> --- a/docs/platform/generic.md
>>>> +++ b/docs/platform/generic.md
>>>> @@ -9,10 +9,9 @@ boards.
>>>>
>>>> By default, the generic FDT platform makes following assumptions:
>>>>
>>>> -1. platform FW_TEXT_START is 0x80000000
>>>> -2. platform features are default
>>>> -3. platform stack size is default
>>>> -4. platform has no quirks or work-arounds
>>>> +1. platform features are default
>>>> +2. platform stack size is default
>>>> +3. platform has no quirks or work-arounds
>>>>
>>>> The above assumptions (except 1) can be overridden by adding special platform
>>>> callbacks which will be called based on FDT root node compatible string.
>>>> @@ -33,10 +32,6 @@ Users of the generic FDT platform will have to ensure that:
>>>> To build the platform-specific library and firmware images, provide the
>>>> *PLATFORM=generic* parameter to the top level `make` command.
>>>>
>>>> -For custom FW_TEXT_START, we can build the platform-specific library and
>>>> -firmware images by passing *PLATFORM=generic FW_TEXT_START=<custom_text_start>*
>>>> -parameter to the top level `make` command.
>>>> -
>>>> Platform Options
>>>> ----------------
>>>>
>>>> diff --git a/firmware/fw_base.S b/firmware/fw_base.S
>>>> index b950c0b..9f995a2 100644
>>>> --- a/firmware/fw_base.S
>>>> +++ b/firmware/fw_base.S
>>>> @@ -53,9 +53,7 @@ _try_lottery:
>>>> bnez a6, _wait_for_boot_hart
>>>>
>>>> /* relocate the global table content */
>>>> - li t0, FW_TEXT_START /* link start */
>>>> - lla t1, _fw_start /* load start */
>>>> - sub t2, t1, t0 /* load offset */
>>>> + lla t2, _fw_start
>>>> lla t0, __rel_dyn_start
>>>> lla t1, __rel_dyn_end
>>>> beq t0, t1, _relocate_done
>>>> diff --git a/firmware/fw_base.ldS b/firmware/fw_base.ldS
>>>> index fb47984..9d11db5 100644
>>>> --- a/firmware/fw_base.ldS
>>>> +++ b/firmware/fw_base.ldS
>>>> @@ -7,8 +7,7 @@
>>>> * Anup Patel <anup.patel at wdc.com>
>>>> */
>>>>
>>>> - . = FW_TEXT_START;
>>>> - /* Don't add any section between FW_TEXT_START and _fw_start */
>>>> + /* Don't add any section before _fw_start */
>>>> PROVIDE(_fw_start = .);
>>>>
>>>> . = ALIGN(0x1000); /* Need this to create proper sections */
>>>> diff --git a/firmware/fw_payload.elf.ldS b/firmware/fw_payload.elf.ldS
>>>> index f1a544b..94e1ac6 100644
>>>> --- a/firmware/fw_payload.elf.ldS
>>>> +++ b/firmware/fw_payload.elf.ldS
>>>> @@ -15,7 +15,7 @@ SECTIONS
>>>> #include "fw_base.ldS"
>>>>
>>>> #ifdef FW_PAYLOAD_OFFSET
>>>> - . = FW_TEXT_START + FW_PAYLOAD_OFFSET;
>>>> + . = FW_PAYLOAD_OFFSET;
>>>> #else
>>>> . = ALIGN(FW_PAYLOAD_ALIGN);
>>>> #endif
>>>> diff --git a/firmware/objects.mk b/firmware/objects.mk
>>>> index e6b364b..a51ff65 100644
>>>> --- a/firmware/objects.mk
>>>> +++ b/firmware/objects.mk
>>>> @@ -13,10 +13,6 @@ firmware-cflags-y +=
>>>> firmware-asflags-y +=
>>>> firmware-ldflags-y +=
>>>>
>>>> -ifdef FW_TEXT_START
>>>> -firmware-genflags-y += -DFW_TEXT_START=$(FW_TEXT_START)
>>>> -endif
>>>> -
>>>> ifdef FW_FDT_PATH
>>>> firmware-genflags-y += -DFW_FDT_PATH=\"$(FW_FDT_PATH)\"
>>>> ifdef FW_FDT_PADDING
>>>> diff --git a/firmware/payloads/test.elf.ldS b/firmware/payloads/test.elf.ldS
>>>> index 2328a1b..08e008f 100644
>>>> --- a/firmware/payloads/test.elf.ldS
>>>> +++ b/firmware/payloads/test.elf.ldS
>>>> @@ -13,7 +13,7 @@ ENTRY(_start)
>>>> SECTIONS
>>>> {
>>>> #ifdef FW_PAYLOAD_OFFSET
>>>> - . = FW_TEXT_START + FW_PAYLOAD_OFFSET;
>>>> + . = FW_PAYLOAD_OFFSET;
>>>> #else
>>>> . = ALIGN(FW_PAYLOAD_ALIGN);
>>>> #endif
>>>> diff --git a/platform/fpga/ariane/objects.mk b/platform/fpga/ariane/objects.mk
>>>> index 83581ac..d1177f4 100644
>>>> --- a/platform/fpga/ariane/objects.mk
>>>> +++ b/platform/fpga/ariane/objects.mk
>>>> @@ -17,7 +17,6 @@ platform-objs-y += platform.o
>>>> PLATFORM_RISCV_XLEN = 64
>>>>
>>>> # Blobs to build
>>>> -FW_TEXT_START=0x80000000
>>>> FW_JUMP=n
>>>>
>>>> ifeq ($(PLATFORM_RISCV_XLEN), 32)
>>>> diff --git a/platform/fpga/openpiton/objects.mk b/platform/fpga/openpiton/objects.mk
>>>> index c8c345a..1a0ce0c 100644
>>>> --- a/platform/fpga/openpiton/objects.mk
>>>> +++ b/platform/fpga/openpiton/objects.mk
>>>> @@ -16,7 +16,6 @@ platform-objs-y += platform.o
>>>> PLATFORM_RISCV_XLEN = 64
>>>>
>>>> # Blobs to build
>>>> -FW_TEXT_START=0x80000000
>>>> FW_JUMP=n
>>>>
>>>> ifeq ($(PLATFORM_RISCV_XLEN), 32)
>>>> diff --git a/platform/generic/objects.mk b/platform/generic/objects.mk
>>>> index 85aa723..c215935 100644
>>>> --- a/platform/generic/objects.mk
>>>> +++ b/platform/generic/objects.mk
>>>> @@ -15,14 +15,13 @@ platform-ldflags-y =
>>>>
>>>> # Command for platform specific "make run"
>>>> platform-runcmd = qemu-system-riscv$(PLATFORM_RISCV_XLEN) -M virt -m 256M \
>>>> - -nographic -bios $(build_dir)/platform/generic/firmware/fw_payload.elf
>>>> + -nographic -bios $(build_dir)/platform/generic/firmware/fw_payload.bin
>>>>
>>>> # Objects to build
>>>> platform-objs-y += platform.o
>>>> platform-objs-y += platform_override_modules.o
>>>>
>>>> # Blobs to build
>>>> -FW_TEXT_START=0x80000000
>>>> FW_DYNAMIC=y
>>>> FW_JUMP=y
>>>> ifeq ($(PLATFORM_RISCV_XLEN), 32)
>>>> diff --git a/platform/kendryte/k210/objects.mk b/platform/kendryte/k210/objects.mk
>>>> index 1bfb898..efac3d2 100644
>>>> --- a/platform/kendryte/k210/objects.mk
>>>> +++ b/platform/kendryte/k210/objects.mk
>>>> @@ -21,6 +21,5 @@ platform-varprefix-k210.o = dt_k210
>>>> platform-padding-k210.o = 2048
>>>>
>>>> # Blobs to build
>>>> -FW_TEXT_START=0x80000000
>>>> FW_PAYLOAD=y
>>>> FW_PAYLOAD_ALIGN=0x1000
>>>> diff --git a/platform/nuclei/ux600/objects.mk b/platform/nuclei/ux600/objects.mk
>>>> index 7c429e0..2753a7f 100644
>>>> --- a/platform/nuclei/ux600/objects.mk
>>>> +++ b/platform/nuclei/ux600/objects.mk
>>>> @@ -22,7 +22,6 @@ platform-runcmd = xl_spike \
>>>> platform-objs-y += platform.o
>>>>
>>>> # Blobs to build
>>>> -FW_TEXT_START=0xA0000000
>>>> FW_DYNAMIC=y
>>>> FW_JUMP=y
>>>>
>>>> diff --git a/platform/template/objects.mk b/platform/template/objects.mk
>>>> index b143cbc..f240a55 100644
>>>> --- a/platform/template/objects.mk
>>>> +++ b/platform/template/objects.mk
>>>> @@ -41,9 +41,6 @@ platform-objs-y += platform.o
>>>> #
>>>> # platform-objs-y += <dt file name>.o
>>>>
>>>> -# Firmware load address configuration. This is mandatory.
>>>> -FW_TEXT_START=0x80000000
>>>> -
>>>> # Optional parameter for path to external FDT
>>>> # FW_FDT_PATH="path to platform flattened device tree file"
>>>>
>>>> @@ -69,8 +66,7 @@ FW_JUMP=<y|n>
>>>> # endif
>>>> # FW_JUMP_FDT_OFFSET=0x2200000
>>>> #
>>>> -# You can use fixed address for jump firmware as an alternative option,
>>>> -# but this may fail when setting wrong FW_TEXT_START. Use with caution.
>>>> +# You can use fixed address for jump firmware as an alternative option.
>>>> # SBI will prefer "<X>_ADDR" if both "<X>_ADDR" and "<X>_OFFSET" are
>>>> # defined
>>>> # ifeq ($(PLATFORM_RISCV_XLEN), 32)
>>>> @@ -97,8 +93,7 @@ endif
>>>> # FW_PAYLOAD_PATH="path to next boot stage binary image file"
>>>> # FW_PAYLOAD_FDT_OFFSET=0x2200000
>>>> #
>>>> -# You can use fixed address for payload firmware as an alternative option,
>>>> -# but this may fail when setting wrong FW_TEXT_START. Use with caution.
>>>> +# You can use fixed address for payload firmware as an alternative option.
>>>> # SBI will prefer "FW_PAYLOAD_FDT_ADDR" if both "FW_PAYLOAD_FDT_OFFSET"
>>>> # and "FW_PAYLOAD_FDT_ADDR" are defined.
>>>> # FW_PAYLOAD_FDT_ADDR=0x82200000
>>>> --
>>>> 2.43.0
More information about the opensbi
mailing list