[8/10] binfmt/elf: Load FDPIC modules through the ELF loader

[casaroli]

depends-on: [/pull/20130]

Summary

[5/10] #19942 and [6/10] #20089 are merged: an FDPIC object is placed, bound and relocated. [7/10] #20130 lets the firmware call back into it. This makes exec() of one work.

The task needs the module's data base in its PIC base register. binfmt builds a D-Space for any object with a GOT, taking the base from the .got address; an FDPIC object names it in DT_PLTGOT instead, which the loader has already translated. The two are the same idea reached by different routes, and both are what up_initial_state() installs.

Constructors are not binfmt's business. A module carries its own crt0, which walks .init_array on the task that runs the module and then calls main, so they run in the module's own context with its own data base. For a module arriving through dlopen(), libelf_insert() walks the array instead, entering each entry through fdpic_invoke(), because a descriptor resolved on the calling task carries the wrong base.

The read-only segment of a module that executes in place is held by a filesystem pin. The load takes it, and the module owns it from the point where nothing can fail any more; it is given back when the task exits. The pin is held through a reference to the file rather than a descriptor, because the descriptor belongs to the task that called the loader while the release happens on another.

libelf_remove() and libelf_uninit() give back what an FDPIC module holds: the pin, and the writable segment. The read-only one is media rather than an allocation and must not be freed.

Impact

Behind CONFIG_FDPIC, which defaults off. With it off, binfmt is what it was.

Testing

mps3-an547:picostest builds with CONFIG_FDPIC off on master alone, and with it on, on top of [7/10].

The constructor and destructor loops enter the module through fdpic_invoke(), which [7/10] defines. They are compiled only under CONFIG_FDPIC, so with it off this one still stands alone on master.

tools/checkpatch.sh -c -u -m -g passes.

Review

[6/10] is merged. This is the point where [7/10] and this one first have to work together, so it wants reading beside [7/10].

[github-actions]

MemBrowse Memory Report

arduino-mega2560

• flash: .data +16 B (+0.0%, 65,144 B / 262,144 B, total: 25% used)
• sram: .data +16 B (+0.6%, 2,718 B / 8,192 B, total: 33% used)

esp32-devkitc

• ROM: .dram0.data +16 B, .flash.rodata +12 B (+0.0%, 124,932 B / 4,194,272 B, total: 3% used)
• dram0_0_seg: .dram0.data +16 B (+0.2%, 9,932 B / 180,736 B, total: 5% used)
• drom0_0_seg: .flash.rodata +12 B (+0.1%, 13,472 B / 4,194,272 B, total: 0% used)

hifive1-revb

• flash: .data +8 B, .text +16 B (+0.0%, 83,748 B / 4,194,304 B, total: 2% used)
• sram: .bss -16 B, .data +8 B (-0.2%, 3,956 B / 16,384 B, total: 24% used)

mirtoo

• kseg0_progmem: .data +4 B, .text +12 B (+0.0%, 67,788 B / 131,072 B, total: 52% used)
• kseg1_datamem: .data +4 B (+0.2%, 2,588 B / 32,256 B, total: 8% used)

qemu-armv8a

• Code: .rodata +16 B, .text.qsort -896 B (+0.0%, 344,384 B)
• Data: .data.g_version +2 B (+0.0%, 77,476 B)

qemu-intel64

• Code: .text +16 B (+0.0%, 8,659,785 B)

rx65n-rsk2mb

• RAM: .data +4 B (+0.1%, 6,590 B / 262,144 B, total: 3% used)
• ROM: .rodata +16 B (+0.0%, 86,912 B / 2,097,152 B, total: 4% used)

s698pm-dkit

• Code: .text +256 B (+0.1%, 367,936 B)

stm32-nucleo-f103rb

• flash: .text +16 B (+0.0%, 34,572 B / 131,072 B, total: 26% used)

[github-actions]

🔗 Cross-repo PR dependencies

The read-only Build run reported the following dependent PR(s) and fetched head SHA(s):

• [6/10] libs/libc/machine/arm: Relocate FDPIC function descriptors #20089 @ 74836bc
• [7/10] libc, sched: Resolve FDPIC descriptors at module callback entry points #20130 @ 75cc4ef

CI run: https://github.com/apache/nuttx/actions/runs/34775786256

[github-actions]

🔗 Cross-repo PR dependencies

The read-only Build run reported the following dependent PR(s) and fetched head SHA(s):

• [7/10] libc, sched: Resolve FDPIC descriptors at module callback entry points #20130 @ 3426186

CI run: https://github.com/apache/nuttx/actions/runs/34957863761

[github-actions]

🔗 Cross-repo PR dependencies

The read-only Build run reported the following dependent PR(s) and fetched head SHA(s):

• [7/10] libc, sched: Resolve FDPIC descriptors at module callback entry points #20130 @ 98f9b2a

CI run: https://github.com/apache/nuttx/actions/runs/35060360729