Shuichiro URATA <ur@a-r.org> makes kernel softfloat emulation code.
http://www.a-r.org/~ur/softfloat1116.diff.gz
is Patch for
sys/arch/mips/conf/files.mips
sys/arch/mips/mips/fp.S
sys/arch/mips/mips/fpemu.c
sys/arch/mips/mips/genassym.cf
sys/arch/mips/mips/locore.S
sys/arch/mips/mips/mips_machdep.c
sys/arch/mips/mips/process_machdep.c
sys/arch/mips/mips/trap.c
sys/arch/mips/mips/vm_machdep.c
After apply this patch,pmax package binary works on hpcmips!
Use flags (formerly access_type) to set page reference/modified status.
Don't use the PG_CLEAN flag from the UVM when checking to see if a
writeable page has been marked as modified.
When updating page status to modified from the UTLBmiss handler, set
the referenced bit in addition to the modified bit.
Page mod/ref status is stored in the pv header, and needs to be copied to
the following entry when removing the head entry, otherwise the status
will be lost (oops!).
Move the common MIPS3 cache flush into pmap_remove_pv() and eliminate the
unnecessary testing of the return value when only compiled for MIPS1.
If the pv entry had the cache inhibited, and we remove the last cache index
alias conflict, restore caching on the mappings for that entry.
Eliminate possible extra cache flushing inherited from the pica pmap: it
was doing the flush when the head entry was being removed - not just the last
entry. Now the flush is done only when the last mapping has been removed.
Also make sure the secondary cache gets flushed [MIPS3 cache flushing needs
to be re-thought/re-done someday].
Update comment for pmap_remove_pv() to reflect these changes.
original debug output was printing the argument to pmap_create(), but
pmap_create() no longer has an argument. The incorrect change now prints
an un-initialized pointer. Change to just print out the function name.
cache. With secondary cache, the CPU will detect cache coherency errors
and the Virtual Coherency Exception handler will flush the appropriate
cache lines to maintain cache coherency. This allows much better
performance than inhibiting the cache for the entire page. This is
very noticable when shared library mappings occur with incompatible
mappings, since there's a very likely chance the mappings will remain
for long periods of time. Systems without secondary cache will still
have the cache inhibited, so there will still be performance issues if
shared libraries don't get mmaped() on correct memory alignments.
This fixes the current problems on DECstations using the R4x00 getting
coredumped programs.
Support VR4100.
Support 16KB page.
Support CPU without FPU.
Fix virtual alias problem(physio() case).
[new options]
options MIPS3_4100 /* VR4100 core */
options MIPS_16K_PAGE /* enable kernel support for 16k pages */
options SOFTFLOAT /* No FPU; avoid touching FPU registers */
big-endian. i386, pc532 and vax still include <machine/byte_swap.h>
and define macros for the {n,h}to{h,n}*() functions. mips also
defines some endian-independent assembly-code aliases for unaligned
memory accesses.
in kernel mode after master interrupt (MIPS_SR_INT_IE) disabled. Tons
of appreciation for Noriyuki Soda and Masanari Tsubai for almost full
time help to spot of the issue.
- remove "need-flag" for mac68k esp driver, as it is not used in anywhere
and conflicts with IPsec ESP header.
This should be the only MD change in IPv6 support, except kernel config file.
Very sorry if you have any compilation problem with it (I believe it is okay).
If your favorite arch is not included in here, please add a
call to ip6intr() from softintr handle.
either endian. Not a perfect solution which would be revealed on
a certain condition when va_arg() is applied to magical struct
arguments passed by value. format_bytes() is now saved. With the
help from Noriyuki Soda and Masanari Tsubai.
TLBUpdate() routine is used for dual purposes. In TLBmod case, just ok
to call 'tlbwi' (as designed). Result in saving of extraneous execution
path. MIPS1 only this moment.
(correct name, vax?) replacing mips1_TLBFlushAddr and mips1_TLBUpdate,
respectively. New codes always use current ASID holded in EntryHi
register. In most occations, the register already contains a necessary
value before (re-)written, ugh. 'sva | asid' ops for their arguments are
now verbose, to be removed when MIPS3 side changes are done.
'pm_asid' member of 'pmap' structure is assigned a new value after
uvmspace_alloc() provides afresh pmap.
- ASID generation number 0 is not a reserved value anymore.
processor is one of processors with no 'referenced bit' nor 'modified bit'
processor machinary. Those functions are implemented combining two
hardware bits, 'dirty bit' and 'valid bit', with TLBmod exception handler.
per process user spaces, replacing mips1_TBLFlush(). This reserves
kernel space TLB entries when TLBPID generation number about to wrap.
- Correct comments a bit, nuke unused routines.
initialization code.
- Abandon mips_init_proc0() and do the 4 lines straightly in MD mach_init().
- Restore a block of code accidentally lost in prevous commit.
- Change the term 'tlbpid' to a MIPS3 nomenclature 'asid'.
- Hide PTE size exposures by symbolic names in locore.S
the child inherits the stack pointer from the parent (traditional
behavior). Like the signal stack, the stack area is secified as
a low address and a size; machine-dependent code accounts for stack
direction.
This is required for clone(2).
_splraise, _spllower, _splset, _splget, _setsoftintr, _clrsoftintr, _splnone.
They manipulate MIPS processor's 8 interrupt sources and are used
as building blocks for NetBSD spl(9) kernel interface. Note that
MIPS processor doesn't enforce inclusive 'interrupt levels' found
in other processors, then the hierarchal nature of IPL must be
implemented by composing MIPS processor interrupt masks appropriately.
With the simplest target port in which small number of devices are
independently assigned with 6 external interrupt signal lines,
spl(9) kernel interface will be implemented with #define's of
processor interrupt controls mentioned above. In more general
cases, in which target computers have many devices and 'system
registers' indicating pending interrupt sources at any moment,
spl(9) will be implemented with more complex machinary manipulating
processor interrupts and system registers in target port dependent
ways.
- Nuke unused code and reorder locore definitions. XXX Following
routines will be replaced with C language version; setrunqueue,
remrunqueue, switchfpregs, savefpregs, MachFPInterrupt.
- ALIAS() is not needed, use XLEAF() or XNESTED() instead
- use AENT() instead of .aent
- _END_LABEL() is not needed (and was wrong)
- define ALEAF(), NLEAF(), NON_LEAF(), NNON_LEAF() by
XLEAF(), LEAF_NONPROFILE(), NESTED(), NESTED_NONPROFILE()
* Map the message buffer with access_type = VM_PROT_READ|VM_PROT_WRITE `just
because'.
* Map the file system buffers with access_type = VM_PROT_READ|VM_PROT_WRITE to
avoid possible problems with pagemove().
* Do not use VM_PROT_EXEC with either of the above.
* Map pages for /dev/mem with access_type = prot. Also, DO NOT use
pmap_kenter() for this, as we DO NOT want to lose modification information.
* Map pages in dumpsys() with VM_PROT_READ.
* Map pages in m68k mappedcopyin()/mappedcopyout() and writeback() with
access_type = prot.
* For now, bus_dma*(), pmap_map(), vmapbuf(), and similar functions still use
access_type = 0. This should probably be revisited.
allocated from a pool, and the MIPS and Alpha use KSEG to map pool
pages. So, mb_map wasn't actually being used. Saves around 4MB of
kernel virtual address space in a typical configuration.
Garbage-collect the related VM_MBUF_SIZE constant.
pointer to just underneath _start. Allows bootblocks (or other kernel
calling programs) to pass more than 4 arguments if it sets up the stack
properly.
The 4kB figure may be overkill - on the pmax it only needs to be a dozen
of so bytes.
tlbmod exception panics. This appears to have been incorrect even in the pica
pmap.c, which I pulled in for the pmax. Fixes Xcfbpmax crashes on R40x0
systems with sfb framebuffer.
Fix resident_count tracking so it doesn't go negative. UVM will change a
page mapping without removing the original mapping. That was causing a
problem early on when getting UVM working on the mips, and the explicit
pmap_remove() done by pmap_enter() was decrementing the resident_count.
Do the pmap_remove() first, then check if the current mapping is not valid
and increment the resident_count.
the *initial* stack pointer. The proc0 stack is being used by proc0 (i.e.
uvm_scheduler). This had been working because switch_exit() wasn't
overwriting any critical areas of the stack - unless kernel profiling
was turned on.
on 64 bit register that's not correctly signed extended. The R4x00 support
works again on DECstations. A similar change to the XTLBMiss handler probably
needs to be made, but I have not done that since I am unable to test any
changes to that.
Also re-order a couple of instructions to allow for delay with mfc0.
- Fix an error in mips3_cpu_switch_resume(); ASID was not set correctly.
- Remove global variable 'curpcb' reference in mips1_proc_trampoline().
- Restore 'cpuregs.h'.
* Add the correct version of locore_mips1.S [ See previous revision for
description of changes ]
* Use defopt'ed MIPS3_L2CACHE_ABSENT in mips_machdep.c and pmap.c
to avoid generating extraneous code.
* GC pmap_set_referenced in pmap.c
minor of libc and the major of libutil). For little-endian architectures
merge the bnswap() assembly versions with nto* and hton* using symbols
aliasing. Use symbol renaming for the bswap function in this case to avoid
namespace pollution.
Declare bswap* in machine/bswap.h, not machine/endian.h. For little-endian
machines, common code for inline macros go in machine/byte_swap.h
Sync libkern with libc.
Adjust #include in kernel sources for machine/bswap.h.
Affected files:
include/mips_param.h, include/pcb.h,
mips/locore_mips1.S, mips/locore_mips3.S,
mips/mips_machdep.c, mips/vm_machdep.c
Issue:
So far, NetBSD/mips has not successfully got rid of fixed-address
kernel stack. USPACE (two 4KB pages) of each process has two distinct
KSEG2 addresses, both refer to a single physical storage; one address
for fixed range [ UADDR .. KERNELSTACK ), and another for "normal" KSEG2
address which was allocated by kernel memory manager and unique to each
others of processes.
"Doubly mapped" USPACE complicates context switch. Both address ranges
have to be managed with a special care of "wired" TLB entries which
are never replaced until next context switch to ensure no TLB miss for
USPACE access. It's equally crumbersome that MIPS processor's cache
machinary gets be confused about USPACE contents because there are two
distinct KSEG2 addresses to manipulate one physical storage.
Solution:
Purge KERNELSTACK constant for kernel stack pointer and replace
it with process unique values. Kernel stack bottom is located at
'curproc->p_addr + USPACE'. Context switch is simplified as it unloads
half of TLB hardwiring burden. It just manages the unique KSEG2 address
of each USPACE to be wired. As the side effect, switch_exit() has no
MIPS processor ISA dependent code anymore. It switchs kernel stack to
proc0's USPACE which has KSEG0 address and no need of TLB entry.
* Extensive use of 'genassym.cf'
To hide target port dependent and/or processor register size dependent
constants from assembler routines, 'genassym.cf' now has an extentive
set of definitions for various constants and offset values of
structural objects. This change will contribute possible NetBSD/mips64
portability too.
* Separation and rename of locore_r2000/_r4000.S
Those files are now indepedent standalones from locore.S to ease
maintainance works, and renamed to match MIPS processor ISA version.
* Changes in kernel mode exception handlers
Kernel mode exception handlers hold exception contexts by pushing a
certain set of register values on stack for resuming kernel mode
processing. This context is now represented with 'struct trapframe',
which is smaller than full scale (user mode) exception context 'struct
frame'. Stack consumption of kernel mode exception services is now
similar to 4.4BSD/mips.
* Relocation of exception frame
User mode context 'struct frame' is moved to the very bottom of kernel
stack at 'curproc->p_addr + USPACE - sizeof(struct frame)' This change
saves a bit of instructions on every return to user processes as it
eliminates reference to global variable 'curpcb' each time.
* Refurblished DDB backtrace routine
It's a growing concern to maintain stacktrace() code correctly. It
could be simplified by enforcing special arrangements for some of
obscure locore routines which violate usual coding conventions.
New backtrace code searchs for certain instructions peculiar to any of
function tails. Specifically, "jr ra" for normal function returns, "jr
k0" for MIPS1 exception handlers and "eret" for MIPS3 handlers.
* Support for 64-bit safe user code
Affected Files:
${ARCH}/include/pubassym.cf lib/libc/arch/mips/gen/*setjmp*
include/setjmp.h mips/include/[lots] mips/mips/[lots]
Solution:
We define macros REG_L/REG_S and SZREG for loading and storing
registers and for the size of registers. The exact meaning
of these is controlled by a macro (currently _MIPS64) which
allows one to treat the registers as either 32-bit or 64-bit.
There are data types mips_reg_t and mips_fpreg_t which represent
the true register sizes, and avoid confusing register_t.
We needed a way to dynamically gen the structure sizes of things
like sigcontext for setjmp.h, so we defined a pubassym.cf for
libc routines like setjmp and longjmp.
NetBSD/mips allows ${ARCH}'s to be defined which preserve
all 64-bits of registers across user context switches. There
are still a few niceties to clean up for kernel mode context
switches.
* Support for QED 52xx processors
Affected Files:
mips/locore_mips3.S mips/pmap.c include/locore.h
Issue:
The QED 52xx family of processors are targeted at low cost
embedded systems, (i.e. CPUs ~$30) for systems like routers, printers,
etc. We have added preliminary support for some of the idiosyncrasies
of this processor, e.g. no L2 cache, etc. More work needs to be
done here because with a modest 2-way L1 cache, some of the rampant
flushing has significant performance implications. However,
it doesn't crash, which is a start.
Solution:
A routine for flushing the cache based on virtual addresses was added;
a routine which deals with the two-way set associativity of the
5230 L1 cache was added, accomodations to 5230's instruction hazards
were added.
* TLB Miss code for mips3/mips4 processors cleaned up significantly.
Affected Files:
mips/locore_mips3.S mips/vm_machdep.c include/locore.h mips_machdep.c
Issue:
The TLB Miss handler exceeded the allowed size, which wasn't
a problem because there was no handler for when the processor
was in 64-bit mode. The handler for invalid TLB exceptions
also appears to have much vestigial code, which made it
difficult to understand.
Solution:
Use the XCONTEXT register to store a pointer to the segment
map table, this coupled with removing some dead code allows
the handlers to fit.
decoupled from long or int or long long. Define macros in asm.h to facilitate
choosing these on a port by port basis.
* Create <machine/pubassym.h> mechanism to allow jmp_buf structure size
to be calculated at system build time.
* Define _MIPS_BSD_SIM macro which specifies what calling style is appropritae
for the architecture. For 64-bit oriented systems set the Status Register
to allow 64-bit instructions.
* Remove UADDR related macros because kernel U structure is now mapped
normally.
* Separate cpu.h into cpu.h and cpuarch.h to keep things neat.
* Add support for QED 52xx processors
- Don't make a reference of curproc when it has NULL value. It causes
double fault upon a fatal panic ocation.
- Macro FETCH_INSTRUCTION() took a value of address 0.
-
UVM relies on pmap modules keeping track of modified/referenced bits
after a page has been removed from all mappings. So *dont* clear
PV_REFERENCED or PV_MODIFIED flags in pmap_remove().
either OP_JR function code or *OP_JALR* function code (not OP_JAL opcode).
insn_unconditional_flow_transfer() was to read an unintialized variable.
Those MD DDB routines seems not useful work so far.
- returned EOPNOTSUPP rather than -1.
- no check for negative offset.
many of these fix potential security problems in these drivers.
XXX XXX XXX
the d_mmap cdev routine should be changed to have a prototype like:
paddr_t (*d_mmap) __P((dev_t, off_t, int));
by someone!
Track page modification status in the PV entry like the alpha, and let
pmap_is_modified() return current status back to the VM system. UVM now
works reliably.
Garbage collect the old pmap_attribute[] stuff.
- cpu_set_kpc() now takes void *arg third argument, passed to the
entry point.
- cpu_fork() allows parent to be non-curproc iff parent is proc0.
When forking non-curproc, assume its state has already been saved.
- Adjust various pieces of machine-dependent code to account of all of this.
* commit isapnpvar.h changes required for ARC to support plain isa.
* fixup mistake over mips/include/cpuregs.h.
* mips/mips_machdep.c:
set L2 cache-size for arc, cleanup use of L2cache present
vs L2 cache-size variables. check for no L2 cache on kernels
configured to require one. misc cleanups.
* mips/mpis/trap.c: more locore stack-traceback label cleanup.
XXX Locore callbacks for mips3, mips4, r4600 cacheflush need more work.
Adds (most) support for ARC platform to port-independent mips code.
Some changes (e.g., clean up of overlapping CPU/FPU ids) inspired by
comparison to the OpenBSD 2.1 codebase of Soda's ARC port.
Open issues:
* Still no support for r4600 or mipsIV CPUs with two-way L1 cache.
Code derived from Per Fogelstrom's OpenBSD source doesn't work
on mips3 pmaxes with L2 cache.
* Still some port-specific #ifdefs, for interrupt enable and
pmax L2 cache-size. Needs more thought, but overlaps with
work-in-progress by Tohru and Tsubai on spl()s and related stuff.
u-area in machine-dependent code. Instead, call exit2() to schedule
the reaper to free them for us, once it is safe to do so (i.e. we are
no longer running on the dead proc's vmspace and stack).
- bump cpu_model[] length as the longest name occupies over 30 characters.
- place machine_arch[] beside machine[] for clearity.
- nuke useless #include directives.
- small scale cleanup in vm_machdep.c
address on 2 architectures anyhow. Also, move the definition of the `label_t'
type inside _KERNEL protection, since it is specific to the in-kernel
setjmp()/longjmp() implementations.
as with user-land programs, include files are installed by each directory
in the tree that has includes to install. (This allows more flexibility
as to what gets installed, makes 'partial installs' easier, and gives us
more options as to which machines' includes get installed at any given
time.) The old SYS_INCLUDES={symlinks,copies} behaviours are _both_
still supported, though at least one bug in the 'symlinks' case is
fixed by this change. Include files can't be build before installation,
so directories that have includes as targets (e.g. dev/pci) have to move
those targets into a different Makefile.
there is one. The Mach VM system seems to take care of this, so it
hasn't knowingly caused a problem. UVM does change mappings without
removing the current mapping, and will pmap_page_protect() hangs
if pmap_enter() doesn't remove the previous mapping.
simonb