This commit removes all traces of Minix segments (the text/data/stack
memory map abstraction in the kernel) and significance of Intel segments
(hardware segments like CS, DS that add offsets to all addressing before
page table translation). This ultimately simplifies the memory layout
and addressing and makes the same layout possible on non-Intel
architectures.
There are only two types of addresses in the world now: virtual
and physical; even the kernel and processes have the same virtual
address space. Kernel and user processes can be distinguished at a
glance as processes won't use 0xF0000000 and above.
No static pre-allocated memory sizes exist any more.
Changes to booting:
. The pre_init.c leaves the kernel and modules exactly as
they were left by the bootloader in physical memory
. The kernel starts running using physical addressing,
loaded at a fixed location given in its linker script by the
bootloader. All code and data in this phase are linked to
this fixed low location.
. It makes a bootstrap pagetable to map itself to a
fixed high location (also in linker script) and jumps to
the high address. All code and data then use this high addressing.
. All code/data symbols linked at the low addresses is prefixed by
an objcopy step with __k_unpaged_*, so that that code cannot
reference highly-linked symbols (which aren't valid yet) or vice
versa (symbols that aren't valid any more).
. The two addressing modes are separated in the linker script by
collecting the unpaged_*.o objects and linking them with low
addresses, and linking the rest high. Some objects are linked
twice, once low and once high.
. The bootstrap phase passes a lot of information (e.g. free memory
list, physical location of the modules, etc.) using the kinfo
struct.
. After this bootstrap the low-linked part is freed.
. The kernel maps in VM into the bootstrap page table so that VM can
begin executing. Its first job is to make page tables for all other
boot processes. So VM runs before RS, and RS gets a fully dynamic,
VM-managed address space. VM gets its privilege info from RS as usual
but that happens after RS starts running.
. Both the kernel loading VM and VM organizing boot processes happen
using the libexec logic. This removes the last reason for VM to
still know much about exec() and vm/exec.c is gone.
Further Implementation:
. All segments are based at 0 and have a 4 GB limit.
. The kernel is mapped in at the top of the virtual address
space so as not to constrain the user processes.
. Processes do not use segments from the LDT at all; there are
no segments in the LDT any more, so no LLDT is needed.
. The Minix segments T/D/S are gone and so none of the
user-space or in-kernel copy functions use them. The copy
functions use a process endpoint of NONE to realize it's
a physical address, virtual otherwise.
. The umap call only makes sense to translate a virtual address
to a physical address now.
. Segments-related calls like newmap and alloc_segments are gone.
. All segments-related translation in VM is gone (vir2map etc).
. Initialization in VM is simpler as no moving around is necessary.
. VM and all other boot processes can be linked wherever they wish
and will be mapped in at the right location by the kernel and VM
respectively.
Other changes:
. The multiboot code is less special: it does not use mb_print
for its diagnostics any more but uses printf() as normal, saving
the output into the diagnostics buffer, only printing to the
screen using the direct print functions if a panic() occurs.
. The multiboot code uses the flexible 'free memory map list'
style to receive the list of free memory if available.
. The kernel determines the memory layout of the processes to
a degree: it tells VM where the kernel starts and ends and
where the kernel wants the top of the process to be. VM then
uses this entire range, i.e. the stack is right at the top,
and mmap()ped bits of memory are placed below that downwards,
and the break grows upwards.
Other Consequences:
. Every process gets its own page table as address spaces
can't be separated any more by segments.
. As all segments are 0-based, there is no distinction between
virtual and linear addresses, nor between userspace and
kernel addresses.
. Less work is done when context switching, leading to a net
performance increase. (8% faster on my machine for 'make servers'.)
. The layout and configuration of the GDT makes sysenter and syscall
possible.
. new mode for sys_memset: include process so memset can be
done in physical or virtual address space.
. add a mode to mmap() that lets a process allocate uninitialized
memory.
. this allows an exec()er (RS, VFS, etc.) to request uninitialized
memory from VM and selectively clear the ranges that don't come
from a file, leaving no uninitialized memory left for the process
to see.
. use callbacks for clearing the process, clearing memory in the
process, and copying into the process; so that the libexec code
can be used from rs, vfs, and in the future, kernel (to load vm)
and vm (to load boot-time processes)
Previously, user processes could cause a kernel panic upon FPU state
restore, by passing bogus FPU state to the kernel (through e.g.
sigreturn). With this patch, the process is now sent a SIGFPE signal
instead.
. only use for single-page invalidations initially
. shows tiny but statistically significant performance
improvement; will be more helpful in certain VM debug
modes
- flush TLB of processes only if the page tables has been changed and
the page tables of this process are already loaded on this cpu which
means that there might be stale entries in TLB. Until now SMP was
always flushing TLB to make sure everything is consistent.
- the Intel architecture cycle counter (performance counter) does not
count when the CPU is idle therefore we use busy loop instead of
halting the cpu when there is nothing to schedule
- the downside is that handling interrupts may be accounted as idle
time if a sample is taken before we get out of the nested trap and
pick a new process
- this makes sure that each process always run with updated TLB
- this is the simplest way how to achieve the consistency. As it means
significant performace degradation when not require, this is nto the
final solution and will be refined
- APs configure local timers
- while configuring local APIC timer the CPUs fiddle with the interrupt
handlers. As the interrupt table is shared the BSP must not run
- to isolate execution inside kernel we use a big kernel lock
implemented as a spinlock
- the lock is acquired asap after entering kernel mode and released as
late as possible. Only one CPU as a time can execute the core kernel
code
- measurement son real hw show that the overhead of this lock is close
to 0% of kernel time for the currnet system
- the overhead of this lock may be as high as 45% of kernel time in
virtual machines depending on the ratio between physical CPUs
available and emulated CPUs. The performance degradation is
significant
- kernel detects CPUs by searching ACPI tables for local apic nodes
- each CPU has its own TSS that points to its own stack. All cpus boot
on the same boot stack (in sequence) but switch to its private stack
as soon as they can.
- final booting code in main() placed in bsp_finish_booting() which is
executed only after the BSP switches to its final stack
- apic functions to send startup interrupts
- assembler functions to handle CPU features not needed for single cpu
mode like memory barries, HT detection etc.
- new files kernel/smp.[ch], kernel/arch/i386/arch_smp.c and
kernel/arch/i386/include/arch_smp.h
- 16-bit trampoline code for the APs. It is executed by each AP after
receiving startup IPIs it brings up the CPUs to 32bit mode and let
them spin in an infinite loop so they don't do any damage.
- implementation of kernel spinlock
- CONFIG_SMP and CONFIG_MAX_CPUS set by the build system
- most global variables carry information which is specific to the
local CPU and each CPU must have its own copy
- cpu local variable must be declared in cpulocal.h between
DECLARE_CPULOCAL_START and DECLARE_CPULOCAL_END markers using
DECLARE_CPULOCAL macro
- to access the cpu local data the provided macros must be used
get_cpu_var(cpu, name)
get_cpu_var_ptr(cpu, name)
get_cpulocal_var(name)
get_cpulocal_var_ptr(name)
- using this macros makes future changes in the implementation
possible
- switching to ELF will make the declaration of cpu local data much
simpler, e.g.
CPULOCAL int blah;
anywhere in the kernel source code
- kernel turns on IO APICs if no_apic is _not_ set or is equal 0
- pci driver must use the acpi driver to setup IRQ routing otherwise
the system cannot work correctly except systems like KVM that use
only legacy (E)ISA IRQs 0-15
There seems to have been a broken assumption in the fpu context
restoring code. It restores the context of the running process, without
guarantee that the current process is the one that will be scheduled.
This caused fpu saving for a different process to be triggered without
fpu hardware being enabled, causing an fpu exception in the kernel. This
practically only shows up with DEBUG_RACE on. Fix my thruby+me.
The fix
. is to only set the fpu-in-use-by-this-process flag in the
exception handler, and then take care of fpu restoring when
actually returning to userspace
And the patch
. translates fpu saving and restoring to c in arch_system.c,
getting rid of a juicy chunk of assembly
. makes osfxsr_feature private to arch_system.c
. removes most of the arch dependent code from do_sigsend
-Makefile updates
-Update mkdep
-Build fixes/warning cleanups for some programs
-Restore leading underscores on global syms in kernel asm files
-Increase ramdisk size
