is little or no reason to create a swap pager for small mmap's. The
vm_map_insert code will automatically create a swap pager if the object
becomes too large. This fix, per a request from phk.
and then never accept for sending packet from upper layer anymore
(i.e. ping -f )
Reviewed by: David Greenman <dg@root.com>
Submitted by: amurai@spec.co.jp
performance issues.
1) The pmap module has had too many inlines, and so the
object file is simply bigger than it needs to be.
Some common code is also merged into subroutines.
2) Removal of some *evil* PHYS_TO_VM_PAGE macro calls.
Unfortunately, a few have needed to be added also.
The removal caused the need for more vm_page_lookups.
I added lookup hints to minimize the need for the
page table lookup operations.
3) Removal of some bogus performance improvements, that
mostly made the code more complex (tracking individual
page table page updates unnecessarily). Those improvements
actually hurt 386 processors perf (not that people who
worry about perf use 386 processors anymore :-)).
4) Changed pv queue manipulations/structures to be TAILQ's.
5) The pv queue code has had some performance problems since
day one. Some significant scalability issues are resolved
by threading the pv entries from the pmap AND the physical
address instead of just the physical address. This makes
certain pmap operations run much faster. This does
not affect most micro-benchmarks, but should help loaded system
performance *significantly*. DG helped and came up with most
of the solution for this one.
6) Most if not all pmap bit operations follow the pattern:
pmap_test_bit();
pmap_clear_bit();
That made for twice the necessary pv list traversal. The
pmap interface now supports only pmap_tc_bit type operations:
pmap_[test/clear]_modified, pmap_[test/clear]_referenced.
Additionally, the modified routine now takes a vm_page_t arg
instead of a phys address. This eliminates a PHYS_TO_VM_PAGE
operation.
7) Several rewrites of routines that contain redundant code to
use common routines, so that there is a greater likelihood of
keeping the cache footprint smaller.
blocked in a write() while waiting for the output to drain, sleep only
for tp->t_timeout, not forever. This only seems to happen when there is
either a modem lockup holding the hardware flow control down, or due to
some problem in the driver with processes attempting to write after the
modem has hung up (eg: elm, tf).
when attepmting to add certain types of routes. This problem
only manifested itself in the presence of unconfigured point-to-point
interfaces.
Noticed by: Chuck Cranor <chuck@maria.wustl.edu>
Saves about 280 butes of source per driver, 56 bytes in object size
and another 56 bytes moves from data to bss.
No functional change intended nor expected.
GENERIC should be about one k smaller now :-)
The old system had the misfeature that the only policy it could implement
was tail-drop; the new IF_ENQ_DROP macro/function makes it possible
to implement more sophisticated queueing policies on a system-wide
basis. No code actually uses this yet (although on my machine
I have converted the ethernet and (polled) loopback to use it).
#ifdef DIAGNOSTIC case, and a warning only otherwise.
People who want them to break into the debugger can always set the
breakpoint explicitly. The existing behaviour was a misfeature from
the beginning, in the (wrong) assumption that the SCSI controller must
always be of essential importance to the entire system.
make it more intelligible, improve the partially bogus locking, and
allow for a ``quick re-acquiration'' from a pending release of timer 0
that happened ``recently'', so it was not processed yet by clkintr().
This latter modification now finally allows to play XBoing over
pcaudio without losing sounds or getting complaints. ;-) (XBoing
opens/writes/closes the sound device all over the day.)
Correct locking for sysbeep().
Extensively (:-) reviewed by: bde
of this patch, which had not actually been reviewed by Joerg or Paul!
(I'll better stop committing files after midnight ...)
I'm now commiting the latest code, which has been reported to work.
Minor correction to the previous commit message for this file:
The first PCI Lance in a system will get a name of lnc1, the second
will be known as lnc2 and so on. An arbitrary number of cards is
supported in a system ...
before attaching. Without this fix, 3c579(EISA) never make
any H/W inturrupt.
Reviewed by: "Justin T. Gibbs" <gibbs@freefall.freebsd.org>, nao@sbl.cl.nec.co.jp and owner-current on mailing list ;-)
Submitted by: amurai@spec.co.jp, nao@sbl.cl.nec.co.jp
This code applies to several systems with integrated Ethernet
chip, for example from HP or Compaq. It should also support
PCI Ethernet cards based on the AMD PCI Lance chip.
This code has been reviewed (visually) by Paul Richards and
tested (using an ISA Lance board) by Joerg Wunsch.
Since the parameters to nearly each and every single function
had to be changed (generally from unit number to lnc_soft*),
there is some potential for buglets having crept in ...
BEWARE: If you had lnc0 configured to have the ISA probe find
your PCI Lance, then it should now be found by the PCI probe,
and should be automatically configured as pci1 (!!! note the "1").
Reviewed by: paul, joerg
rick@snowhite.cis.uoguelph.ca:
1. Clear B_NEEDCOMMIT in nfs_write to make sure that dirty data is
correctly send to the server. If a buffer was dirtied when it was in
the B_DELWRI+B_NEEDCOMMIT state, the state of the buffer was left
unchanged and when the buffer was later cleaned, just a commit rpc was
made to the server to complete the previous write. Clearing
B_NEEDCOMMIT ensures that another write is made to the server.
2. If a server returned a server (for whatever reason) returned an
answer to a write RPC that implied that fewer bytes than requested
were written, bad things would happen.
3. The setattr operation passed on the atime in stead of the mtime to
the server. The fix is trivial.
4. XIDs always started at 0, but this caused some servers (older DEC
OSF/1 3.0 so I've been told) who had very long-lasting XID caches to
get confused if, after a reboot of a BSD client, RPCs came in with a
XID that had in the past been used before from that client. Patch is
to use the current time in seconds as a starting point for XIDs. The
patch below is not perfect, because it requires the root fs to be
mounted first. This is because of the check BSD systems do, comparing
FS time to system time.
Reviewed by: Bruce Evans, Terry Lambert.
Obtained from: frank@fwi.uva.nl (Frank van der Linden) via rick@snowhite.cis.uoguelph.ca
(author's explaination):
Bit 15 is the flag to request a transmit complete interrupt. The
driver was apparently written to minimize interrupts, and if not for a
3-COM design quirk, everything would be just ducky.
Prior to loading the outbound packet into the FIFO, the driver checks
to see if there's enough space to contain the packet. If not, the
driver requests a transmit-available interrupt when there is
sufficient room. Unfortunately, the card is continuing to process the
prior FIFO, and by the time the driver sets the threshold for a
transmit available interrupt, the space is already available. When
this occurs, the 3COM card ignores the interrupt request, and the
driver is hung waiting for an interrupt that will never occur.
There's probably a more elegant solution, but requesting the transmit
complete interrupt was the easiest to implement. An alternative fix
might be to check free FIFO space again, after requesting the transmit
available interrupt, but I haven't bothered pursuing this. Since the
patch, my 3C590 (PCI, same FIFO interface as 3C509) has been rock
solid.
Submitted by: mevans@candle.com (Mike Evans)
pulled up already. This bug can cause the first packet from a source
to a group to be corrupted when it is delivered to a process listening
on the mrouter.
2.1.5-RELEASE). This will obviously be set "for real" closer to the time.
(some ports use this to differentiate the two branches /dev/kmem kernel
architectures. This exact same procedure happened in November last year
for the 2.1 RELEASE as well.)
