of the 61 bits available within the region for virtual addressing. Since
there's no good way for us to map out the gap in the virtual address space,
limit KVA to the architectural minimum implemented address bits. This still
gives us 1 petabyte of KVA, so no worries.
use the PBVM. This eliminates the implied hardcoding of the
physical address at which the kernel needs to be loaded. Using the
PBVM makes it possible to load the kernel irrespective of the
physical memory organization and allows us to replicate kernel text
on NUMA machines.
While here, reduce the direct-mapped page size to the kernel's
page size so that we can support memory attributes better.
services or PAL procedures. The new implementation is based on
specific functions that are known to be called in certain scenarios
only. This in particular fixes the PAL call to obtain information
about translation registers. In general, the new implementation does
not bank on virtual addresses being direct-mapped and will work when
the kernel uses PBVM.
When new scenarios need to be supported, new functions are added if
the existing functions cannot be changed to handle the new scenario.
If a single generic implementation is possible, it will become clear
in due time.
While here, change bootinfo to a pointer type in anticipation of
future development.
1. The PBVM is in region 4, so if we want to make use of it, we
need region 4 freed up.
2. Region 4 and above cannot be represented by an off_t by virtue
of that type being signed. This is problematic for truss(1),
ktrace(1) and other such programs.
Add support for Pre-Boot Virtual Memory (PBVM) to the loader.
PBVM allows us to link the kernel at a fixed virtual address without
having to make any assumptions about the physical memory layout. On
the SGI Altix 350 for example, there's no usuable physical memory
below 192GB. Also, the PBVM allows us to control better where we're
going to physically load the kernel and its modules so that we can
make sure we load the kernel in memory that's close to the BSP.
The PBVM is managed by a simple page table. The minimum size of the
page table is 4KB (EFI page size) and the maximum is currently set
to 1MB. A page in the PBVM is 64KB, as that's the maximum alignment
one can specify in a linker script. The bottom line is that PBVM is
between 64KB and 8GB in size.
The loader maps the PBVM page table at a fixed virtual address and
using a single translations. The PBVM itself is also mapped using a
single translation for a maximum of 32MB.
While here, increase the heap in the EFI loader from 512KB to 2MB
and set the stage for supporting relocatable modules.
The compiler seems to assume it's a 32-bit integral and rounding to the
page size using the standard expression (((u_long)(x) + mask) & ~mask),
results in a 32-bit value. Dropping the 'U' suffix is enough to have the
compiler treat the expression as a 64-bit integral.
handlers.
o Put the IVT in its own section and keep the supporting code close.
o Make sure the VHPT is sized so that it can be mapped using a single
translation.
o Map the PAL code and VHPT with a translation that has the right size.
Assume the platform has a PAL code size that can be mapped with a
single translations.
o Pass the pointer to the bootinfo structure as an argument to ia64_init().
o Get rid of LOG2_ID_PAGE_SIZE and IA64_ID_PAGE_SIZE. It was used to map
the regions 6 & 7 and was as large as possible. The problem is that we
can't support memory attributes easily if the granuratity is not a page.
We need to support memory attributes because the new USB stack violates
the BUS_DMA(9) interface.
o Update some comments...
NOTE: this is broken for SMP kernels, because the AP startup code hasn't
been updated yet.
o The bootinfo structure is now a virtual pointer.
o Replace VM_MAX_ADDRESS with VM_MAXUSER_ADDRESS and redefine
VM_MAX_ADDRESS as the maximum address possible (~0UL).
o Since we're not using direct-mapped translations, switching
to physical addressing is less trivial. Reserve the boot stack
for running in physical mode and special-case the EFI call,
as we're still on the boot stack.
o Region 4 belongs to the kernel now, not process space.
o Move the backing store in the top half of region 0 now that
region 4 is re-assigned to be part of the kernel.
o De-emphasize VM_MAX_ADDRESS. It's really not used anywhere and probably
means something different than the limit for process address space (we
have VM_MAXUSER_ADDRESS for that).
o Exclude the gateway page from VM_MAXUSER_ADDRESS (i.e. make it the same
as VM_MAX_ADDRESS).
now it uses a very dumb first-touch allocation policy. This will change in
the future.
- Each architecture indicates the maximum number of supported memory domains
via a new VM_NDOMAIN parameter in <machine/vmparam.h>.
- Each cpu now has a PCPU_GET(domain) member to indicate the memory domain
a CPU belongs to. Domain values are dense and numbered from 0.
- When a platform supports multiple domains, the default freelist
(VM_FREELIST_DEFAULT) is split up into N freelists, one for each domain.
The MD code is required to populate an array of mem_affinity structures.
Each entry in the array defines a range of memory (start and end) and a
domain for the range. Multiple entries may be present for a single
domain. The list is terminated by an entry where all fields are zero.
This array of structures is used to split up phys_avail[] regions that
fall in VM_FREELIST_DEFAULT into per-domain freelists.
- Each memory domain has a separate lookup-array of freelists that is
used when fulfulling a physical memory allocation. Right now the
per-domain freelists are listed in a round-robin order for each domain.
In the future a table such as the ACPI SLIT table may be used to order
the per-domain lookup lists based on the penalty for each memory domain
relative to a specific domain. The lookup lists may be examined via a
new vm.phys.lookup_lists sysctl.
- The first-touch policy is implemented by using PCPU_GET(domain) to
pick a lookup list when allocating memory.
Reviewed by: alc
o Eliminate IA64_PHYS_TO_RR6 and change all places where the macro is used
by calling either bus_space_map() or pmap_mapdev().
o Implement bus_space_map() in terms of pmap_mapdev() and implement
bus_space_unmap() in terms of pmap_unmapdev().
o Have ia64_pib hold the uncached virtual address of the processor interrupt
block throughout the kernel's life and access the elements of the PIB
through this structure pointer.
This is a non-functional change with the exception of using ia64_ld1() and
ia64_st8() to write to the PIB. We were still using assignments, for which
the compiler generates semaphore reads -- which cause undefined behaviour
for uncacheable memory. Note also that the memory barriers in ipi_send() are
critical for proper functioning.
With all the mapping of uncached memory done by pmap_mapdev(), we can keep
track of the translations and wire them in the CPU. This then eliminates
the need to reserve a whole region for uncached I/O and it eliminates
translation traps for device I/O accesses.
ways:
(1) Cached pages are no longer kept in the object's resident page
splay tree and memq. Instead, they are kept in a separate per-object
splay tree of cached pages. However, access to this new per-object
splay tree is synchronized by the _free_ page queues lock, not to be
confused with the heavily contended page queues lock. Consequently, a
cached page can be reclaimed by vm_page_alloc(9) without acquiring the
object's lock or the page queues lock.
This solves a problem independently reported by tegge@ and Isilon.
Specifically, they observed the page daemon consuming a great deal of
CPU time because of pages bouncing back and forth between the cache
queue (PQ_CACHE) and the inactive queue (PQ_INACTIVE). The source of
this problem turned out to be a deadlock avoidance strategy employed
when selecting a cached page to reclaim in vm_page_select_cache().
However, the root cause was really that reclaiming a cached page
required the acquisition of an object lock while the page queues lock
was already held. Thus, this change addresses the problem at its
root, by eliminating the need to acquire the object's lock.
Moreover, keeping cached pages in the object's primary splay tree and
memq was, in effect, optimizing for the uncommon case. Cached pages
are reclaimed far, far more often than they are reactivated. Instead,
this change makes reclamation cheaper, especially in terms of
synchronization overhead, and reactivation more expensive, because
reactivated pages will have to be reentered into the object's primary
splay tree and memq.
(2) Cached pages are now stored alongside free pages in the physical
memory allocator's buddy queues, increasing the likelihood that large
allocations of contiguous physical memory (i.e., superpages) will
succeed.
Finally, as a result of this change long-standing restrictions on when
and where a cached page can be reclaimed and returned by
vm_page_alloc(9) are eliminated. Specifically, calls to
vm_page_alloc(9) specifying VM_ALLOC_INTERRUPT can now reclaim and
return a formerly cached page. Consequently, a call to malloc(9)
specifying M_NOWAIT is less likely to fail.
Discussed with: many over the course of the summer, including jeff@,
Justin Husted @ Isilon, peter@, tegge@
Tested by: an earlier version by kris@
Approved by: re (kensmith)
VM_PHYSSEG_SPARSE depending on whether the physical address space is
densely or sparsely populated with memory. The effect of this
definition is to determine which of two implementations of
vm_page_array and PHYS_TO_VM_PAGE() is used. The legacy
implementation is obtained by defining VM_PHYSSEG_DENSE, and a new
implementation that trades off time for space is obtained by defining
VM_PHYSSEG_SPARSE. For now, all architectures except for ia64 and
sparc64 define VM_PHYSSEG_DENSE. Defining VM_PHYSSEG_SPARSE on ia64
allows the entirety of my Itanium 2's memory to be used. Previously,
only the first 1 GB could be used. Defining VM_PHYSSEG_SPARSE on
sparc64 allows USIIIi-based systems to boot without crashing.
This change is a combination of Nathan Whitehorn's patch and my own
work in perforce.
Discussed with: kmacy, marius, Nathan Whitehorn
PR: 112194
vm.kmem_size_min. Useful when using ZFS to make sure that vm.kmem size will
be at least 256mb (for example) without forcing a particular value via vm.kmem_size.
Approved by: njl (mentor)
Reviewed by: alc
by libguile that needs to know the base of the RSE backing store. We
currently do not export the fixed address to userland by means of a
sysctl so user code needs to hardcode it for now. This will be revisited
later.
The RSE backing store is now at the bottom of region 4. The memory stack
is at the top of region 4. This means that the whole region is usable
for the stacks, giving a 61-bit stack space.
Port: lang/guile (depended of x11/gnome2)
latter is a kernel option for IA64_ID_PAGE_SHIFT, which in turn
determines IA64_ID_PAGE_MASK and IA64_ID_PAGE_SIZE.
The constants are used instead of the literal hardcoding (in its
various forms) of the size of the direct mappings created in region
6 and 7. The default and probably only workable size is still 256M,
but for kicks we use 128M for LINT.
prime objectives are:
o Implement a syscall path based on the epc inststruction (see
sys/ia64/ia64/syscall.s).
o Revisit the places were we need to save and restore registers
and define those contexts in terms of the register sets (see
sys/ia64/include/_regset.h).
Secundairy objectives:
o Remove the requirement to use contigmalloc for kernel stacks.
o Better handling of the high FP registers for SMP systems.
o Switch to the new cpu_switch() and cpu_throw() semantics.
o Add a good unwinder to reconstruct contexts for the rare
cases we need to (see sys/contrib/ia64/libuwx)
Many files are affected by this change. Functionally it boils
down to:
o The EPC syscall doesn't preserve registers it does not need
to preserve and places the arguments differently on the stack.
This affects libc and truss.
o The address of the kernel page directory (kptdir) had to
be unstaticized for use by the nested TLB fault handler.
The name has been changed to ia64_kptdir to avoid conflicts.
The renaming affects libkvm.
o The trapframe only contains the special registers and the
scratch registers. For syscalls using the EPC syscall path
no scratch registers are saved. This affects all places where
the trapframe is accessed. Most notably the unaligned access
handler, the signal delivery code and the debugger.
o Context switching only partly saves the special registers
and the preserved registers. This affects cpu_switch() and
triggered the move to the new semantics, which additionally
affects cpu_throw().
o The high FP registers are either in the PCB or on some
CPU. context switching for them is done lazily. This affects
trap().
o The mcontext has room for all registers, but not all of them
have to be defined in all cases. This mostly affects signal
delivery code now. The *context syscalls are as of yet still
unimplemented.
Many details went into the removal of the requirement to use
contigmalloc for kernel stacks. The details are mostly CPU
specific and limited to exception_save() and exception_restore().
The few places where we create, destroy or switch stacks were
mostly simplified by not having to construct physical addresses
and additionally saving the virtual addresses for later use.
Besides more efficient context saving and restoring, which of
course yields a noticable speedup, this also fixes the dreaded
SMP bootup problem as a side-effect. The details of which are
still not fully understood.
This change includes all the necessary backward compatibility
code to have it handle older userland binaries that use the
break instruction for syscalls. Support for break-based syscalls
has been pessimized in favor of a clean implementation. Due to
the overall better performance of the kernel, this will still
be notived as an improvement if it's noticed at all.
Approved by: re@ (jhb)
- Don't include ia64_cpu.h and cpu.h
- Guard definitions by _NO_NAMESPACE_POLLUTION
- Move definition of KERNBASE to vmparam.h
o Move definitions of IA64_RR_{BASE|MASK} to vmparam.h
o Move definitions of IA64_PHYS_TO_RR{6|7} to vmparam.h
o While here, remove some left-over Alpha references.
disk drivers along with a load of fixes to context switching, fork
handling and a load of other stuff I can't remember now. This takes us as
far as start_init() before it dies. I guess now I will have to finish off
the VM system and syscall handling :-).
not work on any real hardware (or fully work on any simulator). Much more
needs to happen before this is actually functional but its nice to see
the FreeBSD copyright message appear in the ia64 simulator.
