cacheable window on physical memory (KSEG0). On the Sibyte processor
going through the uncacheable window (KSEG1) bypasses both L1 and L2
caches so we may end up with stale contents in the L2 cache.
This also makes it consistent with the rest of the function that
uses cacheable mappings to copy pages.
Approved by: imp (mentor)
fixed-state media with parameters set via hints
and configure MAC accordingly to these parameters.
All the underlying PHY magic is done by boot manager
on startup. At the moment there is no proper way
to make active and control all PHYs simultaneously
from one MII bus and there is no way to associate
incoming/outgoing packet with specific PHY.
original codes (I had changed one by accident)
Also do the pic_ack/pic_delayed_ack after the interrupt
so we clear it. The clock with these changes starts working.
Its off doing a short/long short/long warning but it
now runs.
My NFS mount now works but has the same problem with
sbin/init (errno 8 ENOEXEC) so it panics with no init.
Either this is a problem with my buildworld.. OR its a
yet undiscovered RMI issue.
- Move dpcpu initialization to mips_proc0_init. It's
more appropriate place for it. Besides dpcpu_init
requires pmap module to be initialized and calling it
int pmap.c hangs the system
the mi_startup (or to the last of it).. and
hit a panic after :
uart0: <16550 or compatible> on iodi0
Trap cause = 2 (TLB miss....)
I did have to take the pci bus OUT of the
build to get this far, hit a cache error with
the PCI code in. Interesting thing is the machine
reboots too ;-)
addresses and could modify areas of memory that share the same cache
line at the beginning and at the ending of the buffer. In order to
prevent a data loss we save these chunks in temporary buffer before
invalidation and restore them afer it.
Idea suggested by: cognet
while in kernel mode, and later changing signal mask to block the
signal, was fixed for sigprocmask(2) and ptread_exit(3). The same race
exists for sigreturn(2), setcontext(2) and swapcontext(2) syscalls.
Use kern_sigprocmask() instead of direct manipulation of td_sigmask to
reschedule newly blocked signals, closing the race.
Reviewed by: davidxu
Tested by: pho
MFC after: 1 month
The "_MTC0 v0, COP_0_TLB_HI" is actually incorrect because v0 has not been
initialized at that point. It worked correctly because we subsequently
did the right thing and initialized TLB_HI correctly.
The "li v0, MIPS_KSEG0_START" is redundant because we do exactly the same
thing 2 instructions down.
Approved by: imp (mentor)
SDRAM_ADDR_START, SDRAM_ADDR_END and SDRAM_MEM_SIZE
Instead we now keep a copy of the memory regions enumerated by
platform-specific code and use that to determine whether an address
is cacheable or not.
Approved by: imp (mentor)
the memory or D-cache, depending on the semantics of the platform.
vm_sync_icache() is basically a wrapper around pmap_sync_icache(),
that translates the vm_map_t argumument to pmap_t.
o Introduce pmap_sync_icache() to all PMAP implementation. For powerpc
it replaces the pmap_page_executable() function, added to solve
the I-cache problem in uiomove_fromphys().
o In proc_rwmem() call vm_sync_icache() when writing to a page that
has execute permissions. This assures that when breakpoints are
written, the I-cache will be coherent and the process will actually
hit the breakpoint.
o This also fixes the Book-E PMAP implementation that was missing
necessary locking while trying to deal with the I-cache coherency
in pmap_enter() (read: mmu_booke_enter_locked).
The key property of this change is that the I-cache is made coherent
*after* writes have been done. Doing it in the PMAP layer when adding
or changing a mapping means that the I-cache is made coherent *before*
any writes happen. The difference is key when the I-cache prefetches.
in cpu_switch and use it in stack_trace function later. pcb_regs contains
state of the process stored by exception handler and therefor is not
valid for sleeping processes.
Context info could be obtained from other sources (see below) no only from
td_pcb field
- Do not show a0..a3 values unless they're obtained from the stack. These
are only confirmed values.
- Fix bt command in DDB. Previous implementation used thread's trapframe
structure as a source info for trace unwinding, but this structure
is filled only when exception occurs. Valid register values for sleeping
processes are in pcb_context array. For curthread use pc/sp/ra for current
frame
compile and many of them may disappear. For example
the xlr_boot1_console.c is old code that is ifdef'd out.
I will clean these sorts of things up as I make progress
on the port. So far the only thing I have I think straightened
out is the bits around the interupt handling... and hey that
may be broke ;-)
1) Adds future RMI directories
2) Places intr_machdep.c in specfic files.arch pointing to the generic
intr_machdep.c. This allows us to have an architecture dependant intr_machdep.c
(which we will need for RMI) in the machine specific directory
3) removes intr_machdep.c from files.mips
4) Adds some TARGET_XLR_XLS ifdef's for the machine specific intra_machdep.h. We
may need to look at finding a better place to put this. But first I want to
get this thing compiling.
by looking at the bases used for non-relocatable executables by gnu ld(1),
and adjusting it slightly.
Discussed with: bz
Reviewed by: kan
Tested by: bz (i386, amd64), bsam (linux)
MFC after: some time
first and the native ia32 compat as middle (before other things).
o(ld)brandinfo as well as third party like linux, kfreebsd, etc.
stays on SI_ORDER_ANY coming last.
The reason for this is only to make sure that even in case we would
overflow the MAX_BRANDS sized array, the native FreeBSD brandinfo
would still be there and the system would be operational.
Reviewed by: kib
MFC after: 1 month
like we do in Malta. We may want to look at consolidating things
because *ALL* mips will *ALWAYS* be memory mapped. The only wrinkle
is that the tag may need to be a custom one (see endian issues with
the Atheros port for one example).
former works for both ILP32 and LP64 programming models, while the
latter fails LP64.
# uintpr_t is better, but iirc, we can't pollute the name space to use it
# I likely need to audit all my uintptr_t changes for that issue...
has proven to have a good effect when entering KDB by using a NMI,
but it completely violates all the good rules about interrupts
disabled while holding a spinlock in other occasions. This can be the
cause of deadlocks on events where a normal IPI_STOP is expected.
* Adds an new IPI called IPI_STOP_HARD on all the supported architectures.
This IPI is responsible for sending a stop message among CPUs using a
privileged channel when disponible. In other cases it just does match a
normal IPI_STOP.
Right now the IPI_STOP_HARD functionality uses a NMI on ia32 and amd64
architectures, while on the other has a normal IPI_STOP effect. It is
responsibility of maintainers to eventually implement an hard stop
when necessary and possible.
* Use the new IPI facility in order to implement a new userend SMP kernel
function called stop_cpus_hard(). That is specular to stop_cpu() but
it does use the privileged channel for the stopping facility.
* Let KDB use the newly introduced function stop_cpus_hard() and leave
stop_cpus() for all the other cases
* Disable interrupts on CPU0 when starting the process of APs suspension.
* Style cleanup and comments adding
This patch should fix the reboot/shutdown deadlocks many users are
constantly reporting on mailing lists.
Please don't forget to update your config file with the STOP_NMI
option removal
Reviewed by: jhb
Tested by: pho, bz, rink
Approved by: re (kib)
a device pager (OBJT_DEVICE) object in that it uses fictitious pages to
provide aliases to other memory addresses. The primary difference is that
it uses an sglist(9) to determine the physical addresses for a given offset
into the object instead of invoking the d_mmap() method in a device driver.
Reviewed by: alc
Approved by: re (kensmith)
MFC after: 2 weeks
dependent memory attributes:
Rename vm_cache_mode_t to vm_memattr_t. The new name reflects the
fact that there are machine-dependent memory attributes that have
nothing to do with controlling the cache's behavior.
Introduce vm_object_set_memattr() for setting the default memory
attributes that will be given to an object's pages.
Introduce and use pmap_page_{get,set}_memattr() for getting and
setting a page's machine-dependent memory attributes. Add full
support for these functions on amd64 and i386 and stubs for them on
the other architectures. The function pmap_page_set_memattr() is also
responsible for any other machine-dependent aspects of changing a
page's memory attributes, such as flushing the cache or updating the
direct map. The uses include kmem_alloc_contig(), vm_page_alloc(),
and the device pager:
kmem_alloc_contig() can now be used to allocate kernel memory with
non-default memory attributes on amd64 and i386.
vm_page_alloc() and the device pager will set the memory attributes
for the real or fictitious page according to the object's default
memory attributes.
Update the various pmap functions on amd64 and i386 that map pages to
incorporate each page's memory attributes in the mapping.
Notes: (1) Inherent to this design are safety features that prevent
the specification of inconsistent memory attributes by different
mappings on amd64 and i386. In addition, the device pager provides a
warning when a device driver creates a fictitious page with memory
attributes that are inconsistent with the real page that the
fictitious page is an alias for. (2) Storing the machine-dependent
memory attributes for amd64 and i386 as a dedicated "int" in "struct
md_page" represents a compromise between space efficiency and the ease
of MFCing these changes to RELENG_7.
In collaboration with: jhb
Approved by: re (kib)
that takes into account all limitation to DMA memory (boundaries,
alignment) and implements bounce pages.
- Add BUS_DMASYNC_POSTREAD case to bus_dmamap_sync_buf
(1) fix printf formats.
(2) Prefer FreeBSD's MIPS_PHYS_TO_KSEG0 to hand-rolled one from Cavium.
(3) Mark a few 64-bit cleanliness issues (possible).
(4) Minor formatting fixes.
o add to platforms where it was missing (arm, i386, powerpc, sparc64, sun4v)
o define as "1" on amd64 and i386 where there is no restriction
o make the type returned consistent with ALIGN
o remove _ALIGNED_POINTER
o make associated comments consistent
Reviewed by: bde, imp, marcel
Approved by: re (kensmith)
# Note: Cavium provided a port that has atomics similar to these, but
# that does a syncw; sync; atomic; sync; syncw where we just do the classic
# mips 'atomic' operation (eg ll; frob; sc). It is unclear to me why
# the extra is needed. Since my initial target is one core, I'll defer
# investigation until I bring up multiple cores. syncw is an octeon specific
# instruction.
them to void * first. This neatly solves the "how do I print a
register_t" problem because sizeof(void *) is always the same as
sizeof(register_t), afaik.
this is correct. While registers are 64-bit, n32 is a 32-bit ABI and
lives in a 32-bit world (with explicit 64-bit registers, however).
Change an 8, which was 4 + 4 or sizeof(int) + SZREG to be a simple '4
+ SZREG' to reflect the actual offset of the structure in question.
places. Provide n32/n64 register name defintions. This should have
no effect for the O32 builds that everybody else uses, but should help
make N64 builds possible (lots of other changes are needed for that).
Obtained from: NetBSD (for the regdef.h changes)
IF_ADDR_UNLOCK() across network device drivers when accessing the
per-interface multicast address list, if_multiaddrs. This will
allow us to change the locking strategy without affecting our driver
programming interface or binary interface.
For two wireless drivers, remove unnecessary locking, since they
don't actually access the multicast address list.
Approved by: re (kib)
MFC after: 6 weeks
required by video card drivers. Specifically, this change introduces
vm_cache_mode_t with an appropriate VM_CACHE_DEFAULT definition on all
architectures. In addition, this changes adds a vm_cache_mode_t parameter
to kmem_alloc_contig() and vm_phys_alloc_contig(). These will be the
interfaces for allocating mapped kernel memory and physical memory,
respectively, with non-default cache modes.
In collaboration with: jhb
- Modules and kernel code alike may use DPCPU_DEFINE(),
DPCPU_GET(), DPCPU_SET(), etc. akin to the statically defined
PCPU_*. Requires only one extra instruction more than PCPU_* and is
virtually the same as __thread for builtin and much faster for shared
objects. DPCPU variables can be initialized when defined.
- Modules are supported by relocating the module's per-cpu linker set
over space reserved in the kernel. Modules may fail to load if there
is insufficient space available.
- Track space available for modules with a one-off extent allocator.
Free may block for memory to allocate space for an extent.
Reviewed by: jhb, rwatson, kan, sam, grehan, marius, marcel, stas
trap() function re-enables interrupts if exception happened with
interrupts enabled and therefor status register might be modified
by interrupt filters
atomic_fetchadd_32. Without it gcc would use it as input
register for v and sometimes generate following code for
function call like atomic_fetchadd_32(&(fp)->f_count, -1):
801238b4: 2402ffff li v0,-1
801238b8: c2230018 ll v1,24(s1)
801238bc: 00431021 addu v0,v0,v1
801238c0: e2220018 sc v0,24(s1)
801238c4: 1040fffc beqz v0,801238b8 <dupfdopen+0x2e8>
801238c8: 00000000 nop
Which is definitly wrong because if sc fails v0 is set to 0
and previous value of -1 is overriden hence whole operation
turns to bogus
Yes, this puts things in the wrong place, doesn't compile and is
woefully incomplete. However, it will allow us to more easily track
against the upstream sources without needing to import the entire
Cavium tree under vendor.
This port is based on FreeBSD 7.0 as of April 2007 and the pre-import
MIPS tree (aka mips2), so much work is necessary here.
to prevent race over k0, k1 registers.
- Update interrupts mask in saved status register for
MipsUserIntr and MipsUserGenException. It might be
modified by intr filter or ithread.
hard and soft interrupts
- Do not handle masked interrupts
- Do not write Cause register because most bytes are read-only and
writing the same byte to RW fields are pointless. And in case of
software interrupt utterly wrong
- Get rid of arge_fix_chain, use m_defrag like if_vr
- Rework interrupt handling routine to avoid race that lead
to disabling RX interrupts
- Enable full duplex if requested
- Properly set station MAC address
- Slightly optimize RX loop
- Initialize FILTERMATCH and FILTERMASK registers as linux driver does
- Replace a1 with k1 to while restoring context. a1 was there by mistake,
interrupts are disabled at this point and it's safe to use k0, k1.
This code never was reached beacasue current Status register handling
prevented interrupta from user mode.
write fault or while wiring a mapping that must support write access.
In general, this change should reduce the number of traps that occur for
the purpose of setting the modified bit. More specifically, this change
should prevent traps while holding locks in a sysctl handler. See
kern/kern_sysctl.c revisions 1.168 and 1.195 (svn r192160) for further
details.
Tested by: gonzo
* In arge_attach(), hard reset the MAC blocks before configuring the MAC.
* In arge_reset_dma(), clear pending packet interrupts based off
the hardware counter instead of acking every packet in the ring,
as the hardware counter can exceed the ring size. If the reset
was successful the counters will be zero anyway.
* In arge_encap(), remove an unused variable.
* In arge_tx_locked(), remove redundant setting of the EMPTY flag as
the TX DMA engine sets it for us.
* In arge_intr(), remember to clear the interrupt status bits
relayed from arge_intr_filter().
* Handle RX overflow and TX underflow.
* In arge_tx_intr(), remember to unmask the TX interrupt bits
after processing them.
register increments only every second cycle. The only timing
references for us is Count value. Therefore it's better to convert
frequencies related to it and use them. Besides cleanup this commit
fixes twice more then requested sleep interval problem.
possible future I-cache coherency operation can succeed. On ARM
for example the L1 cache can be (is) virtually mapped, which
means that any I/O that uses temporary mappings will not see the
I-cache made coherent. On ia64 a similar behaviour has been
observed. By flushing the D-cache, execution of binaries backed
by md(4) and/or NFS work reliably.
For Book-E (powerpc), execution over NFS exhibits SIGILL once in
a while as well, though cpu_flush_dcache() hasn't been implemented
yet.
Doing an explicit D-cache flush as part of the non-DMA based I/O
read operation eliminates the need to do it as part of the
I-cache coherency operation itself and as such avoids pessimizing
the DMA-based I/O read operations for which D-cache are already
flushed/invalidated. It also allows future optimizations whereby
the bcopy() followed by the D-cache flush can be integrated in a
single operation, which could be implemented using on-chips DMA
engines, by-passing the D-cache altogether.
Reimplement "kernel_pmap" in the standard way.
Eliminate unused variables. (These are mostly variables that were
discarded by the machine-independent layer after FreeBSD 4.x.)
Properly handle a vm_page_alloc() failure in pmap_init().
Eliminate dead or legacy (FreeBSD 4.x) code.
Eliminate unnecessary page queues locking.
Eliminate some excess white space.
Correct the synchronization of pmap_page_exists_quick().
Tested by: gonzo
a fair number of static data structures, making this an unlikely
option to try to change without also changing source code. [1]
Change default cache line size on ia64, sparc64, and sun4v to 128
bytes, as this was what rtld-elf was already using on those
platforms. [2]
Suggested by: bde [1], jhb [2]
MFC after: 2 weeks
Introduce pmap_try_insert_pv_entry(), a function that conditionally
creates a pv entry if the number of entries is below the high water mark
for pv entries.
Introduce pmap_enter_quick_locked() and use it to reimplement
pmap_enter_object(). The old implementation was broken. For example,
it could block while holding a mutex lock.
Change pmap_enter_quick_locked() to fail rather than wait if it is
unable to allocate a page table page. This prevents a race between
pmap_enter_object() and the page daemon. Specifically, an inactive
page that is a successor to the page that was given to
pmap_enter_quick_locked() might become a cache page while
pmap_enter_quick_locked() waits and later pmap_enter_object() maps
the cache page violating the invariant that cache pages are never
mapped. Similarly, change
pmap_enter_quick_locked() to call pmap_try_insert_pv_entry() rather
than pmap_insert_entry(). Generally speaking,
pmap_enter_quick_locked() is used to create speculative mappings. So,
it should not try hard to allocate memory if free memory is scarce.
Tested by: gonzo
registers with ar71xx_bus_space_reversed. Note, that byte order
of values is handled by drivers. bus_spaces fixes only position
of register in word.
- Replace .hints hack for AR71XX UART with ar71xx_bus_space_reversed.
CACHE_LINE_SIZE constant. These constants are intended to
over-estimate the cache line size, and be used at compile-time
when a run-time tuning alternative isn't appropriate or
available.
Defaults for all architectures are 64 bytes, except powerpc
where it is 128 bytes (used on G5 systems).
MFC after: 2 weeks
Discussed on: arch@
to bus_space structure that defines access methods and hence every bus can
define own accessors. Default space is mips_bus_space_generic. It's a simple
interface to physical memory, values are read with regard to host system
byte order.
1) Move the new field (brand_note) to the end of the Brandinfo structure.
2) Add a new flag BI_BRAND_NOTE that indicates that the brand_note pointer
is valid.
3) Use the brand_note field if the flag BI_BRAND_NOTE is set and as old
modules won't have the flag set, so the new field brand_note would be
ignored.
Suggested by: jhb
Reviewed by: jhb
Approved by: kib (mentor)
MFC after: 6 days
