in_cksum.h required ip.h to be included for struct ip. To be
able to use some general checksum functions like in_addword()
in a non-IPv4 context, limit the (also exported to user space)
IPv4 specific functions to the times, when the ip.h header is
present and IPVERSION is defined (to 4).
We should consider more general checksum (updating) functions
to also allow easier incremental checksum updates in the L3/4
stack and firewalls, as well as ponder further requirements by
certain NIC drivers needing slightly different pseudo values
in offloading cases. Thinking in terms of a better "library".
Sponsored by: The FreeBSD Foundation
Sponsored by: iXsystems
Reviewed by: gnn (as part of the whole)
MFC After: 3 days
1. Define all registers. These definitions are needed to support
the FCM driver for direct-connect NAND.
2. Repurpose lbc_read_reg() and lbc_write_reg() for use by localbus
attached device drivers. Use bus_space functions directly in the
lbc driver itself.
3. Be smarter about programming LAWs and mapping memory. The ranges
defined in the FDT are per bank (= chip select) and since we can
have up to 8 banks, we could easily use more than 8 LAWs or TLB
enrties when per-bank memory ranges need multiple LAWs or TLBs
due to alignment or size constraints.
We now combine all memory ranges into the fewest possible set of
contiguous regions and program the hardware for that. Thus, a
cleverly written FDT with 8 devices may still only need 1 LAW or
1 TLB entry. Note that the memory ranges can be assigned randomly
to the banks. We sort as we build to handle that.
4. Support the FCM when programming the OR register. This is mostly
for documention purposes as we do not have a way to define the
mode for a bank.
5. Remove Semihalf-ism: do not define DEBUG (only to undefine it
again).
FDT does not define all ranges possible for a particular node (e.g.
PCI).
While here, only update the trgt_mem and trgt_io pointers if there's
no error. This avoids that we knowingly write an invalid target (= -1).
for variables that live in the boot page.
o Add bp_trace (yes, it's in the boot page) that gets zeroed before we
try to wake a core and to which the core being woken can write markers
so that we know where the core was in case it doesn't wake up. The
boot code does not yet write markers (too follow).
o Disable the boot page translation to allow the last 4K page to be used
for whatever we please. It would get mapped otherwise.
o Fix kernstart in the case of SMP. The start argument is typically page
aligned due to the alignment requirements that come with having a boot
page. The point of using trunc_page is that we get the actual load
address given that the entry point is immediately following the ELF
headers. In the SMP case this ended up exactly 4K after the load
address. Hence subtracting 1 from start.
exceptions early enough during boot that the kernel will do ithe same.
Use lwsync only when compiling for LP64 and revert to the more proven isync
when compiling for ILP32. Note that in the end (i.e. between revision 222198
and this change) ILP32 changed from using sync to using isync. As per Nathan
the isync is needed to make sure I/O accesses are properly serialized with
locks and isync tends to be more effecient than sync.
While here, undefine __ATOMIC_ACQ and __ATOMIC_REL at the end of the file
so as not to leak their definitions.
Discussed with: nwhitehorn
range operations like pmap_remove() and pmap_protect() as well as allowing
simple operations like pmap_extract() not to involve any global state.
This substantially reduces lock coverages for the global table lock and
improves concurrency.
- Use isync/lwsync unconditionally for acquire/release. Use of isync
guarantees a complete memory barrier, which is important for serialization
of bus space accesses with mutexes on multi-processor systems.
- Go back to using sync as the I/O memory barrier, which solves the same
problem as above with respect to mutex release using lwsync, while not
penalizing non-I/O operations like a return to sync on the atomic release
operations would.
- Place an acquisition barrier around thread lock acquisition in
cpu_switchin().
guarantees on acquire for the tlbie mutex. Conversely, the TLB invalidation
sequence provides guarantees that do not need to be redundantly applied on
release. Roll a small custom lock that is just right. Simultaneously,
convert the SLB tree changes back to lwsync, as changing them to sync
was a misdiagnosis of the tlbie barrier problem this commit actually fixes.
not provide general barriers, but only barriers in the context of the
atomic sequences here. As such, make them private and keep the global
*mb() routines using a variant of sync.
isync to implement read and write barriers, following Appendix B.2 of
Book II of the architecture manual. This provides a 25% speed increase
to fork() on the PowerPC G4.
of sync (lwsync is an alternate encoding of sync on systems that do not
support it, providing graceful fallback). This provides more than an order
of magnitude reduction in the time required to acquire or release a mutex.
MFC after: 2 months
sync performs a strict superset of the functions of eieio, so using both
is redundant. While here, expand bus barriers to all bus_space operations,
since many drivers do not correctly use bus_space_barrier().
In principle, we can also replace sync just with eieio, for a significant
performance increase, but it remains to be seen whether any poorly-written
drivers currently depend on the side effects of sync to properly function.
MFC after: 1 week
the page. This PMAP requires an additional lock besides the PMAP lock
in pmap_extract_and_hold(), which vm_page_pa_tryrelock() did not release.
Suggested by: kib
MFC after: 4 days
for whether the page is physical. On dense phys mem systems (32-bit),
VM_PHYS_TO_PAGE will not return NULL for device memory pages if device
memory is above physical memory even if there is no allocated vm_page.
Attempting to use the returned page could then cause either memory
corruption or a page fault.
usermode context switches (long jumps and ucontext operations). If these
are used across threads, multiple threads can end up with the same TLS base.
Madness will then result.
This makes behavior on PPC match that on x86 systems and on Linux.
MFC after: 10 days
(slightly) different semantics and renaming it prevents a (harmless)
WITNESS warning during bootup for 32-bit kernels on 64-bit CPUs.
MFC after: 5 days
be less ambiguous and more clearly identify what it means. This
attribute is what Intel refers to as UC-, and it's only difference
relative to normal UC memory is that a WC MTRR will override a UC-
PAT entry causing the memory to be treated as WC, whereas a UC PAT
entry will always override the MTRR.
- Remove the VM_MEMATTR_UNCACHED alias from powerpc.
New kernel events can be added at various location for sampling or counting.
This will for example allow easy system profiling whatever the processor is
with known tools like pmcstat(8).
Simultaneous usage of software PMC and hardware PMC is possible, for example
looking at the lock acquire failure, page fault while sampling on
instructions.
Sponsored by: NETASQ
MFC after: 1 month
are are not mapped during ranged operations and reduce the scope of the
tlbie lock only to the actual tlbie instruction instead of the entire
sequence. There are a few more optimization possibilities here as well.
