o Allocate a VHPT per CPU. The VHPT is a hash table that the CPU
uses to look up translations it can't find in the TLB. As such,
the VHPT serves as a level 1 cache (the TLB being a level 0 cache)
and best results are obtained when it's not shared between CPUs.
The collision chain (i.e. the hash bucket) is shared between CPUs,
as all buckets together constitute our collection of PTEs. To
achieve this, the collision chain does not point to the first PTE
in the list anymore, but to a hash bucket head structure. The
head structure contains the pointer to the first PTE in the list,
as well as a mutex to lock the bucket. Thus, each bucket is locked
independently of each other. With at least 1024 buckets in the VHPT,
this provides for sufficiently finei-grained locking to make the
ssolution scalable to large SMP machines.
o Add synchronisation to the lazy FP context switching. We do this
with a seperate per-thread lock. On SMP machines the lazy high FP
context switching without synchronisation caused inconsistent
state, which resulted in a panic. Since the use of the high FP
registers is not common, it's possible that races exist. The ia64
package build has proven to be a good stress test, so this will
get plenty of exercise in the near future.
o Don't use the local ID of the processor we want to send the IPI to
as the argument to ipi_send(). use the struct pcpu pointer instead.
The reason for this is that IPI delivery is unreliable. It has been
observed that sending an IPI to a CPU causes it to receive a stray
external interrupt. As such, we need a way to make the delivery
reliable. The intended solution is to queue requests in the target
CPU's per-CPU structure and use a single IPI to inform the CPU that
there's a new entry in the queue. If that IPI gets lost, the CPU
can check it's queue at any convenient time (such as for each
clock interrupt). This also allows us to send requests to a CPU
without interrupting it, if such would be beneficial.
With these changes SMP is almost working. There are still some random
process crashes and the machine can hang due to having the IPI lost
that deals with the high FP context switch.
The overhead of introducing the hash bucket head structure results
in a performance degradation of about 1% for UP (extra pointer
indirection). This is surprisingly small and is offset by gaining
reasonably/good scalable SMP support.
allocating a VHPT per CPU. Since we don't yet know how many CPUs
are actually in the system at the time we need to allocate the
VHPTs, we allocate for MAXCPU processors. This can result in a
lot of wasted space for 2-way machines. So, for now, limit MAXCPU
to something smaller until we have something more dynamic.
command does, but worse.
o Remove the obscure proc command, because it does what the thread
command does, but not unambigously.
o Move the PID to the extra thread info, where it makes sense and
where it doesn't confuse users. The extra thread info holds some
process information, to which the PID belongs.
o Implement the to_find_new_threads target method by having it call
the target beneath us if we're not using KVM. This makes sure that
new threads are found when using the remote target.
o Fix various core dump scenarios:
- Implement the to_files_info target method. Previously the
'info target' command would cause a NULL pointer dereference.
- Don't assume there's a current thread. We're not initialized
in all cases. This prevents a NULL pointer dereference.
- When we're not ussing KVM, have the to_xfer_memory target
method call the target beneath us. This avoids calling into
KVM with a NULL pointer.
MFC after: 1 week
static.
o Register a function with atexit(3) to close the KVM object if
we have one open.
o Show the unread portion of the kernel's message buffer before
presenting the prompt. It's bound to provide some useful info.
o Don't call kgdb_target() twice. It results in having all threads
listed twice.
MFC after: 1 week
so that libmemstat can be used to view full memory statistics from
kernel core dumps and /dev/mem. This is provided via a new query
function, memstat_kvm_malloc(), which is also automatically invoked
by memstat_kvm_all(). A kvm handle must be passed in.
This will allow malloc(9)-specific code to be removed from vmstat(8).
integer to an unsigned long. This lifts variables like the maximum
number of pages available for shared memory from 2^31 to 2^32 on 32
bit architectures, and from 2^31 to 2^64 on 64 bit architectures.
It should be noted that this changes breaks ABI on 64 bit architectures
because the size of the shmmax, shmmin, shmmni, shmseg and shmall members
of the shminfo structure has changed.
Silence on: current@
when operating in ap mode. Previously we allocated a node from the
station table, sent the frame (using the node), then released the
reference that "held the frame in the table". But while the frame
was in flight the node might be reclaimed which could lead to
problems. The solution is to add an ieee80211_tmp_node routine
that crafts a node that does exist in a table and so isn't ever
reclaimed; it exists only so long as the associated frame is in flight.
MFC after: 5 days
constructing and applying binary patches; in particular, they perform
well (in the sense of constructing small patches) for executable code.
Both portsnap (coming to the base system Real Soon Now) and FreeBSD
Update (coming to the base system a bit later) use bspatch.
This is the same code as the bsdiff-4.2 which has been in the ports
tree (misc/bsdiff) for the past year, with the following exceptions:
1. The license is now the traditional 2-clause BSD;
2. Instead of forking and execing bzip2, the code now uses libbz2; and
3. Some minor changes have been made to fit this code into the base
system (adding $FreeBSD$ tags, putting bsdiff and bspatch into separate
directories, etc.)
This code is rather ugly and has lots of style bugs (mostly because I
wrote it before I had ever heard of style(9)). Some day I'll come
back and clean it up.
Discussed on: freebsd-arch
MFC before: 5.5-RELEASE
Tested by: Several million users (earlier version).
vnode is inactivated), possibly leading to a NULL dereference when
checking if the mount wants knotes to be activated in the VOP hooks.
So, we add a new vnode flag VV_NOKNOTE that is only set in getnewvnode(),
if necessary, and check it when activating knotes.
Since the flags are not erased when a vnode is being held, we can safely
read them.
Reviewed by: kris@
MFC after: 3 days
Previously, we used all info (including -1 or "not present") which would
keep the system from reaching 100% when charging.
Reported by: Eric Anderson
MFC after: 2 days
- Add locked versions of the init() and start() methods.
- Use callout_*() rather than timeout().
- Make the driver lock non-recursive.
- Push down locking in detach() and ioctl().
- Fix the tick routine to bail if the interface has been stopped and use
callout_drain() in detach() after the call to stop().
- Lock the driver lock in the ifmedia handlers.
Tested by: Ketrien I. Saihr-Kesenchedra ketrien at error404.nls.net
MFC after: 1 week
it is running during installations. The vsystem() function does get
used after this happens (e.g. if you manually configure the network
interface and ask it to enable the interface) so it needs to be a
little smarter about whether it uses /stand/sh or /bin/sh.
MFC after: 3 days
