fpu: save each process's x87/SSE registers; keep the kernel off them

Nothing ever saved the FPU and SSE registers on a context switch.
musl's memcpy, the compiler's struct copies and any arithmetic on
doubles keep live values in %xmm, so two processes sharing a CPU wrote
over each other's: two shells counting in a loop on one processor took
a NULL+8 page fault within seconds, at a different rip every time. It
reproduces on kernels from before this week too -- it had always been
there, and every crash nobody could explain was a candidate.

sched() now fxsaves the process leaving the CPU and scheduler()
fxrstors the one arriving. A new process starts from the ABI's state
(FCW 0x37F, MXCSR 0x1F80), fork copies the parent's live registers,
exec resets them, and a signal handler's are put back by rt_sigreturn
so the interrupted code finds what it had.

The kernel was free to use %xmm too -- GCC vectorised struct copies
and memset into 153 instructions of it -- so an interrupt could
clobber the registers of the program it interrupted with no switch at
all. Kernel, lwIP and uACPI objects are built -mgeneral-regs-only now
and the image has none; cmd/ keeps SSE.

Checked under QEMU: three concurrent loops three times over, on 1 CPU
and on 4, no faults; before, every run faulted.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01S1giNvrHTrPMPw7V6x1kQ4
This commit is contained in:
Nexoniarz 2026-10-23 15:02:01 +02:00
commit 0e39e711f1
6 changed files with 69 additions and 0 deletions

View file

@ -425,6 +425,14 @@ CFLAGS = -fno-pic -static -fno-builtin -fno-strict-aliasing -Wall -MD -MP -ggdb
CFLAGS += -ffreestanding -fno-common -nostdlib -I./include $(XFLAGS) $(OPT)
CFLAGS += $(NO_SSP) $(CFLAGS_EXTRA)
# The kernel keeps its hands off the FPU and SSE registers: they belong to
# whichever process is running, and are saved only when it is switched out
# (see "The FPU" in kernel/proc/proc.c). Without this GCC vectorises struct
# copies and memset into %xmm, and an interrupt then scribbles over the
# registers of the program it interrupted. Kernel objects only -- cmd/ is
# built with these CFLAGS too and is welcome to use SSE.
$(BUILD)/kernel/%.o $(BUILD)/lwip/%.o $(BUILD)/uacpi/%.o: CFLAGS += -mgeneral-regs-only
# lwIP's own headers, plus ours (cc.h and lwipopts.h live in kernel/net/lwip).
# Only the network files are compiled with these: putting them in the global
# CFLAGS would let any kernel file include lwIP by accident, and the stack is