move cxgb_lt2.[ch] from NIC to TOE
move most offload functionality from NIC to TOE
factor out all socket and inpcb direct access
factor out access to locking in incpb, pcbinfo, and sockbuf
Add the '-u name' option to the env command, which will completely unset
the given variable name (removing it from the environment, instead of
just setting it to a null value).
Next stage of stdio cleanup: Retire __sFILEX and merge the fields back into
__sFILE. This was supposed to be done in 6.0. Some notes:
- Where possible I restored the various lines to their pre-__sFILEX state.
- Retire INITEXTRA() and just initialize the wchar bits (orientation and
mbstate) explicitly instead. The various places that used INITEXTRA
didn't need the locking fields or _up initialized. (Some places needed
_up to exist and not be off the end of a NULL or garbage pointer, but
they didn't require it to be initialized to a specific value.)
- For now, stdio.h "knows" that pthread_t is a 'struct pthread *' to
avoid namespace pollution of including all the pthread types in stdio.h.
Once we remove all the inlines and make __sFILE private it can go back
to using pthread_t, etc.
- This does not remove any of the inlines currently and does not change
any of the public ABI of 'FILE'.
Change the timeout(9) usage in if_ndis to a callout(9) implementation,
as the former is becoming deprecated and exhibits some extraneous
Giant-locking. The new callout(9) is declared MPSAFE, so it may
improve concurrency.
Convert pcbinfo and inpcb mutexes to rwlocks, and modify macros to
explicitly select write locking for all use of the inpcb mutex.
Update some pcbinfo lock assertions to assert locked rather than
write-locked, although in practice almost all uses of the pcbinfo
rwlock main exclusive, and all instances of inpcb lock acquisition
are exclusive.
This change should introduce (ideally) little functional change.
However, it lays the groundwork for significantly increased
parallelism in the TCP/IP code.
Allow linker_search_symbol_name() to be called with KLD lock held.
The linker_search_symbol_name() function is used by stack_print()
and stack_print() can be called from kernel module unload method.
marius [Thu, 17 Apr 2008 12:38:00 +0000 (12:38 +0000)]
On sparc64 machines with multiple host-PCI-bridges these bridges
have separate configuration spaces so by definition they implement
different PCI domains. Thus change psycho(4) to use PCI domains
instead of reenumerating all PCI busses so they have globally unique
bus numbers and drop support for reenumerating busses in the OFW PCI
code.
According to CVS history reenumeration was also required in order to
get some E450 to boot but given that no other open source kernel
changes the PCI bus numbers assigned by the firmware I believe the
real problem was that the old code used the bus number as the device
number for the PCI busses and unlike most of the other machines the
firmwares of the problematic ones don't use disjoint PCI bus numbers
across the host-PCI-bridges.
sos [Thu, 17 Apr 2008 12:29:35 +0000 (12:29 +0000)]
Go back to preallocating everything possible on init.
This avoids calling busdma in the request processing path which caused a traumatic performance degradation.
Allocation has be postponed to after we know how many devices we possible can have on portmulitpliers to save some space.
jeff [Thu, 17 Apr 2008 09:56:01 +0000 (09:56 +0000)]
- Add a metric to describe how busy a processor has been over the last
two ticks by counting the number of switches and the load when
sched_clock() is called.
- If the busy metric exceeds a threshold allow the idle thread to spin
waiting for new work for a brief period to avoid using IPIs. This
reduces the cost on the sender and receiver as well as reducing wakeup
latency considerably when it works.
jeff [Thu, 17 Apr 2008 04:20:10 +0000 (04:20 +0000)]
- Make SCHED_STATS more generic by adding a wrapper to create the
variables and sysctl nodes.
- In reset walk the children of kern_sched_stats and reset the counters
via the oid_arg1 pointer. This allows us to add arbitrary counters to
the tree and still reset them properly.
- Define a set of switch types to be passed with flags to mi_switch().
These types are named SWT_*. These types correspond to SCHED_STATS
counters and are automatically handled in this way.
- Make the new SWT_ types more specific than the older switch stats.
There are now stats for idle switches, remote idle wakeups, remote
preemption ithreads idling, etc.
- Add switch statistics for ULE's pickcpu algorithm. These stats include
how much migration there is, how often affinity was successful, how
often threads were migrated to the local cpu on wakeup, etc.
Simplify the pmap_zero_page family of functions by making use of
the fact that we have a 1:1 mapping by virtue of the BATs.
Eliminate the now unused moea_rkva_alloc(), moea_pa_map() and
moea_pa_unmap() functions.
Use separate UMA zone for data items allocation. It is a partial
rev. 1.149 rework.
It allows to save several percents of CPU time on SMP by using UMA's
internal per-CPU allocation limits instead of own global variable
each time updated with atomics.
Implement fdopendir(3) by splitting __opendir2() into two parts, the upper part
deals with the usual __opendir2() calls, and the rest part with an interface
translator to expose fdopendir(3) functionality. Manual page was obtained from
kib@'s work for *at(2) system calls.
Move the head of byte-level advisory lock list from the
filesystem-specific vnode data to the struct vnode. Provide the
default implementation for the vop_advlock and vop_advlockasync.
Purge the locks on the vnode reclaim by using the lf_purgelocks().
The default implementation is augmented for the nfs and smbfs.
In the nfs_advlock, push the Giant inside the nfs_dolock.
Before the change, the vop_advlock and vop_advlockasync have taken the
unlocked vnode and dereferenced the fs-private inode data, racing with
with the vnode reclamation due to forced unmount. Now, the vop_getattr
under the shared vnode lock is used to obtain the inode size, and
later, in the lf_advlockasync, after locking the vnode interlock, the
VI_DOOMED flag is checked to prevent an operation on the doomed vnode.
The implementation of the lf_purgelocks() is submitted by dfr.
Add very limited support for the isc-dhclient. It will almostly certaintly
only work if there's just one interface doing dhcp. This version implements
the same logic as the version in the PR, but uses pgrep to be less verbose.
When sending packets directly to the DHCP server, use a socket and send
directly rather than bogusly sending it out as a link layer broadcast
(which fails to be received on some networks).
Several changes breaking netgraph module ABI collected together:
- reorder structures fields (XX_refs) a bit to group fields modified
same time together. According to my tests it gives up to 10%
SMP performance benefit on real workload due to reduced inter-CPU
cache trashing.
- change q_flags from long to int as long is not really needed there and
it's usage with atomics is argued by some people.
- move NGF_WORKQ flag into the separate field q_flags2 as it protected by
queue mutex instead of node writer protection used by the rest of flags.
- move nd_work queue entry to ng_queue structure to which it is more
related and make it STAILQ instead of TAILQ as now it is a classic FIFO.
- remove q_node pointer from ng_queue structure as it is not really needed.
- reimplement item queue using STAILQ instead of own equal implementation.
As soon as BT subsystem has own item queues using ng_item.el_next update
it also.
- change depth field in ng_item from uintptr_t to u_int. It was made
uintptr_t to keep ABI compatibility.
Use genclock for RTC handling. This eliminates the MD versions for
inittodr() and resettodr(). Have nexus double as the clock device,
because it's the firmware that provides RTC services. We could
create a special (pseudo-) device for it, but that wasn't superior
enough to actually do it. Maybe later...
This file shows signs of heavy borrowing from NetBSD's cache.c file,
so credit its authors with contributions to this file. Remove
prototype copyright notice, although one might be warranted if someone
wanted to claim it badly enough.
Add copyright notice and license to reflect origin of most of the
routines in this file. Remove 'place holder' copyright since the
amount that's actually original is small relative to the length of the
file. The contents of this file appear to have originated at DECWRL
by way of NetBSD.
Support and switch to the ULE scheduler:
o Implement IPI_PREEMPT,
o Set td_lock for the thread being switched out,
o For ULE & SMP, loop while td_lock points to blocked_lock for
the thread being switched in,
o Enable ULE by default in GENERIC and SKI,
Don't touch MSI enable bit in RL_CFG2 register. For unknown reason
clearing MSI enable bit for MSI capable hardwares resulted in Tx
problems. MSI enable bit is set only when MSI is requested from
user.
Revision 1.9 changes the delivery mode from the magic constant 0
(i.e. fixed delivery) to SAPIC_DELMODE_LOWPRI. While the commit
log doesn't mention the change in behaviour, it is believed to be
deliberate. In the last 5.5 years this hasn't been a problem. Nor
do I think did it make any difference, but who knows. However, I
do know that it break SMP support for Montecito-based machines.
Switch back to fixed-CPU delivery so that SMP works again. This
gives me some time to look more closely at the problem, as well
as make sure the I-cache validation as it's implemented currently
is sufficient in SMP configurations...
restore(8) does not check for write failure while building two temp
files containing directory and ownership data. If /tmp fills, the
console is blasted with zillions of "file system full" errors, and
restore continues on, even though directory and/or ownership data
has been lost. This is particularly likely to happen when running
from the live CD, which has little /tmp space.
PR: bin/93603, also probably bin/107213
Fix from: Ken Lalonde
sos [Mon, 14 Apr 2008 18:34:24 +0000 (18:34 +0000)]
Fix problem with slave devices.
Fix or rather bring ENOMEM problems back to the state it was before.
Temporarily disable PortMultipliers on AHCI devices.
Protection against errant sender sending a stream
seq number out of order with no missing TSN's (a
cisco box has this problem which will make a ssn
be held forever).
MFC after: 1 week
FreeBSD/mips port. The FreeBSD/mips port targets mips32, mips64,
mips32r2 and mips64r2 (and close relatives) processors. There
presently is support for ADMtek ADM5120, A mips 4Kc in a malta board,
the RB533 routerboard (based on IDT RC32434) and some preliminary
support for sibtye/broadcom designs. Other hardware support will be
forthcomcing.
This port boots multiuser under gxemul emulating the malta board and
also bootstraps on the hardware whose support is forthcoming...
Oleksandr Tymoshenko, Wojciech Koszek, Warner Losh, Olivier Houchard,
Randall Stewert and others that have contributed to the mips2 and/or
mips2-jnpr perforce branches. Juniper contirbuted a generic mips port
late in the life cycle of the misp2 branch. Warner Losh merged the
mips2 and Juniper code bases, and others list above have worked for
the past several months to get to multiuser.
In addition, the mips2 work owe a debt to the trail blazing efforts of
the original mips branch in perforce done by Juli Mallett.
FreeBSD/mips port. The FreeBSD/mips port targets mips32, mips64,
mips32r2 and mips64r2 (and close relatives) processors. There
presently is support for ADMtek ADM5120, A mips 4Kc in a malta board,
the RB533 routerboard (based on IDT RC32434) and some preliminary
support for sibtye/broadcom designs. Other hardware support will be
forthcomcing.
This port boots multiuser under gxemul emulating the malta board and
also bootstraps on the hardware whose support is forthcoming...
Oleksandr Tymoshenko, Wojciech Koszek, Warner Losh, Olivier Houchard,
Randall Stewert and others that have contributed to the mips2 and/or
mips2-jnpr perforce branches. Juniper contirbuted a generic mips port
late in the life cycle of the misp2 branch. Warner Losh merged the
mips2 and Juniper code bases, and others list above have worked for
the past several months to get to multiuser.
In addition, the mips2 work owe a debt to the trail blazing efforts of
the original mips branch in perforce done by Juli Mallett.