- fill in and use ifp->if_softc
- use if_bpf rather than private cookie variables
- change bpf interface to take advantage of this
- call ether_ifattach() directly from Ethernet drivers
- delete kludge in if_attach() that did this indirectly
and rmx_recvpipe. This has no demonstrable effect on performance.
(ttcp reports about 44 Mbit/s for all the buffer sizes I tried between
16384 and 65536.)
pmap_activate since it's not used anymore. Changed cpu_fork so that
it uses one line of inline assembly rather than calling mvesp() to
get the current stack pointer. Removed mvesp() since it is no longer
being used.
fixes for previous version of new pipes from Bruce Evans. This
new version:
Supports more properly the semantics of select (BDE).
Supports "OLD_PIPE" correctly (kern_descrip.c, BDE).
Eliminates incorrect EPIPE returns (bash 'pipe broken' messages.)
Much faster yet, currently tuned relatively conservatively -- but now
gives approx 50% more perf than the new pipes code did originally.
(That was about 50% more perf than the original BSD pipe code.)
Known bugs outstanding:
No support for async io (SIGIO). Will be included soon.
Next to do:
Merge support for FIFOs.
Submitted by: bde
for me, but has gotten a bit flakey in bidirectional parallel port mode.
Fix a bug in bidirectional parallel port transfers, more work is still
needed here (testers welcome).
Minor cleanup.
drives require ST_Q_SNS_HLP, they also wrongly accept a blocksize of
1024 in the first place (for a QIC-150 cartridge), but complain later
about it. The hack is to only probe for 512 for them.
Reorder the entries in st_decide_mode() so that QIC >= 525 is properly
accepted as variable blocksize.
vs unidirectional transfer modes. The kernel handles hardware, user mode
programs shouldn't get in the way.
This cleans up some really ugly grots that I hated too. :-)
Suggested by: Sujal Patel <smpatel@wam.umd.edu>
* this is my unoptimized driver, it works fine, it's not as fast as it
* could be (yet) -- I have yet to merge in ideas from other QuickCam
* developers.
* warning: this user interface is still in flux pending negotiations
* with other quickcam driver authors. It is _not_ compatible with the
* original linux interface due to the fact that it was too restrictive.
that don't announce support for command queues.
SCSI_NCR_DFLT_TAGS can be specified in the kernel config file
and sets the default number of tags per disk drive.
A value of 0 means "no tags".
Minor correction in debug messages: Values from the msg_in
buffer were being printed in the msg_out trace message ...
The code outputs the dc then calls the device specific externalize
routines to fill in the dc_data area. The old code assumed that dc_data
started one byte from the end of the dc, but with the compiler optimizing
alignment and padding, this isn't always the case. Do an explicit
&(dc.dc_data) - &dc. This fixes lsdev -c which must have been broken
for some time.
- Call eisa_registerdev as soon as we have a device match. This allows the
"eisa_add_*" routines to tweak kdc_datalen as the kdc grows and shrinks.
eisaconf.c
- externalize the linked lists that hold our ioaddrs and maddrs.
caused by a different reason):
. #ifndef __FreeBSD__ around check for negative size, FreeBSD size_t is
unsigned
. Disable mirror/parity if interleave size is 0 (i.e., serial concatenation).
when a connection enters the ESTBLS state using T/TCP, then window
scaling wasn't properly handled. The fix is twofold.
1) When the 3WHS completes, make sure that we update our window
scaling state variables.
2) When setting the `virtual advertized window', then make sure
that we do not try to offer a window that is larger than the maximum
window without scaling (TCP_MAXWIN).
Reviewed by: davidg
Reported by: Jerry Chen <chen@Ipsilon.COM>
(1) The reads are always done from the first n/2 disks.
(2) Each write is done twice, to the "data" disk (in the first half) and
the "mirror" disk (in the second half).
ccdbuffer() now takes an extra argument (struct ccdbuf **) and stores
the pointer to ccdbuf in there. In case of a mirrored write, it
allocates and stores two pointers. The "residual" is also doubled
for mirrored writes so that ccdiodone() can correctly tell when all
the writes are done.
clock interrupts.
Keep a 1-in-16 smoothed average of the length of each tick. If the
CPU speed is correctly diagnosed, this should give experienced users
enough information to figure out a more suitable value for `tick'.
fstat() syscall, rather than panic("linux newfstat").
(Note: I've extracted this from a larger set of diffs, I'm confident I've
not missed any dependencies but can't modload it to test it on my system)
feature in the header type register, though it is required by the PCI spec.
This should correctly probe both functions of the Intel 82371FB chip,
without the need for a special case based on the device ID.
1) The calculation didn't account for NMBCLUSTERS, so if a large number of
clusters was specified, it would leave little or no space for kernel
malloc.
2) It was bogusly restricted to v_page_count. This doesn't take into
account the sparseness of the malloc area and would have caused
problems on machines with small amounts of memory. It should probably
instead be changed to set the malloc limit to be constrained by
the amount of memory, but I didn't do this.
since setting up the DMA is too costly. Restructure for efficiency.
Pause the sequencer when a parity error occurs so that the kernel driver
knows during which phase the error was encountered.
. remove some unused variables
. declare worminit() right this time, it's actually extern (?)
. use wormminphys(), now that it's already there (hope i've got this right)
chipset. This does not attempt to do anything special with the timing
on the hope that the BIOS will have done the right thing already. The
actual interface from the wd driver to the new facility is not
implemented yet (this commit being an attempt at prodding someone else
to do it because looking at the wd driver always confuses the h*** out of me).
Also declared worminit() to avoid a compiler warning. Seems that the
other SCSI drivers don't declare XXinit() yet, so perhaps we'd also
create a typedef for these func's.
includes a hack in the probe code: the 82371FB is a multifuction
device, but doesn't properly set the configuration bit which
indicates this. So, we just hard-wire all 82371FBs as multifunction
devices.
This does not actually make the bus-master IDE stuff work, although
if anyone wants to work on that, I have the databooks that tell
how to use it.
The worm driver is alpha-usable! I've stuffed everything that is
needed into the kernel driver, including the logic to select
between different vendor's quirks.
Disclaimer: this has by now _only_ been tested on a heavily tweaked
2.0.5R system. I've done my best to retrofit it into -current, but
i don't have a chance yet to test it in a -current environment.
First attempt at creating devfs entries for sliced devices. Doesn't
quite work yet, so the heart of it is disabled.
Added bdev and cdev args to dsopen().
Create devfs entries in dsopen() and (unsuccessfully) attempt to make
them go away at the right times. DEVFS is #undefed at the start so
that this shouldn't cause problems.
quite work yet, so the heart of it is disabled.
Added bdev and cdev args to dsopen().
drivers:
Fixed device names, links, minor numbers and modes.
wd.c:
Started actually supporting devfs.
diskslice.h:
Added devfs tokens to structs (currently 576 of them per disk! :-().
subr_diskslice.c:
Create devfs entries in dsopen() and (unsuccessfully) attempt to make
them go away at the right times. DEVFS is #undefed at the start so
that this shouldn't cause problems.
fd and wt drivers need bounce buffers, so this normally saves 32K-1K
of kernel memory.
Keep track of which DMA channels are busy. isa_dmadone() must now be
called when DMA has finished or been aborted.
Panic for unallocated and too-small (required) bounce buffers.
fd.c:
There will be new warnings about isa_dmadone() not being called after
DMA has been aborted.
sound/dmabuf.c:
isa_dmadone() needs more parameters than are available, so temporarily
use a new interface isa_dmadone_nobounce() to avoid having to worry
about panics for fake parameters. Untested.
overrun by 2 pages.
Fixed the (unused) values returned from device attach functions.
Fixed checking of unit number in device open functions - don't load bad
pointers or print error messages about the contents of bad pointers.
Removed unused #includes.
asc.c:
Fixed premature setting of flags in ascopen() - copied the better order
in gscopen().
gsc.c:
Fixed conflict handling for drq:
- fail the probe if the configured drq doesn't match the actual drq.
- set the configured drq to match the actual drq in the autoconfig case.
Reviewed by: Nobody; authors didn't respond to mail.
way to avoid crossing a 64K DMA boundary was to specify an alignment
greater than the size even when the alignment didn't matter, and for
sizes larger than a page, this reduced the chance of finding enough
contiguous pages. E.g., allocations of 8K not crossing a 64K boundary
previously had to be allocated on 8K boundaries; now they can be
allocated on any 4K boundary except (64 * n + 60)K.
Fixed bugs in vm_alloc_page_contig():
- the last page wasn't allocated for sizes smaller than a page.
- failures of kmem_alloc_pageable() weren't handled.
Mutated vm_page_alloc_contig() to create a more convenient interface
named contigmalloc(). This is the same as the one in 1.1.5 except
it has `low' and `high' args, and the `alignment' and `boundary'
args are multipliers instead of masks.
of limited utility. In their place, add bunch of pointers
which will eventually be needed by the polled-interrupt scheme we're working
here. (It will probably be a while before the code is written and
committed here.) At the same time, a `void *if_softc' field
was added to the beginning of the structure to make certain driver
writers happier.
The practical upshot of all this is that you need to
recompile utilities such as netstat which manipulate struct ifnet.
This is a really ugly bandaid on the problem, but it works well enough for
'ps -u' to start working again. The problem was caused by the user
address space shrinking by a little bit and the UPAGES being "cast off" to
become a seperate entity rather than being at the top of the process's
vmspace. That optimization was part of John's most recent VM speedups.
Now, rather than decoding the VM space, it merely ensures the pages are
in core and accesses them the same way the ptrace(PT_READ_U..) code does,
ie: off the p->p_addr pointer.
a panic due to an attaempt to allocate a buffer for a terabyte or
so of data when an attempt is made to create sparse data (e.g.
a holey file) more than 1 block past the end of the file.
Note: some other areas of this code need to be looked at,
since they might cause problems when the file size exceeds 2GB,
due to storing results in ints when the computations are being
done with quad sized variables.
Reviewed by: bde
Implement a "variable" directory structure. Files that do not make
sense for the given process do not "appear" and cannot be opened.
For example, "system" processes do not have "file", "regs" or "fpregs",
because they do not have a user area.
"attempt" to fill in the user area of a given process when it is being
accessed via /proc/pid/mem (the user struct is just after
VM_MAXUSER_ADDRESS in the process address space.)
Dont do IO to the U area while it's swapped, hold it in place if possible.
Lock off access to the "ctl" file if it's done a setuid like the other
pseudo-files in there.
PT_ATTACH/PT_DETACH implemented now and fully operational.
PT_{GET|SET}{REGS|FPREFS} implemented now, using code shared with procfs
PT_{READ|WRITE}_{I|D} now uses code shared with procfs
ptrace opcodes now fully permission checked, including ownerships.
doing an operation to the u-area on a swapped process should no longer
panic.
running gdb as root works for me now, where it didn't before.
general cleanup..
Note, that this has some tightening of permissions/access checks etc.
Some of these may be going too far.. In particular, the "owner" of the
traced process is enforced. The process that created or attached to
the traced process is now the only one that can "do" things to it.
This makes it sorta usable, just for my ``proof-of-concept'' Perl
script i've been posting to freebsd-scsi. The driver will be
overhauled further, this is just to provide Jordan with a base to
perform testing of his own with his HP burner. Use entirely at your
own risk, expect a bunch of misburnt CD-R's when using it already in
this very green stage.
Note that by now the driver will only work when the CD-R has already
been in the drive at boot time.
- cpuclass should be cpu_class
- CPUCLASS_I386 should be CPUCLASS_386
(^^ those only show up if you compile for i386)
- two missing prototypes on new functions
- one missing static
to 20000 through 30000. These numbers are used for local IP port numbers
when an explicit address is not specified.
The values are sysctl modifiable under: net.inet.ip.port_{first|last}_auto
These numbers do not overlap with any known server addresses, without going
above 32768 which are "negative" on some other implementations.
20000 through 30000 is 2.5 times larger than the old range, but some have
suggested even that may not be enough... (gasp!) Setting a low address
of 10000 should be plenty.. :-)
Speed up for vfs_bio -- addition of a routine bqrelse to greatly diminish
overhead for merged cache.
Efficiency improvement for vfs_cluster. It used to do alot of redundant
calls to cluster_rbuild.
Correct the ordering for vrele of .text and release of credentials.
Use the selective tlb update for 486/586/P6.
Numerous fixes to the size of objects allocated for files. Additionally,
fixes in the various pagers.
Fixes for proper positioning of vnode_pager_setsize in msdosfs and ext2fs.
Fixes in the swap pager for exhausted resources. The pageout code
will not as readily thrash.
Change the page queue flags (PG_ACTIVE, PG_INACTIVE, PG_FREE, PG_CACHE) into
page queue indices (PQ_ACTIVE, PQ_INACTIVE, PQ_FREE, PQ_CACHE),
thereby improving efficiency of several routines.
Eliminate even more unnecessary vm_page_protect operations.
Significantly speed up process forks.
Make vm_object_page_clean more efficient, thereby eliminating the pause
that happens every 30seconds.
Make sequential clustered writes B_ASYNC instead of B_DELWRI even in the
case of filesystems mounted async.
Fix a panic with busy pages when write clustering is done for non-VMIO
buffers.
Save 112K for SB, 64K for PAS and 64K for MSS.
Since PAS use SB emulation, 176K normally saved for it.
Few minor optimizations added like in Linux driver.
When you start tracker and produce some heavy disk activity,
output interrupts becomes lost and I don't know how to solve it
finally. Newly added code at least allows recovery after timeout.
Fixed BCD declarations. They didn't match their definitions...
libkern.h, bcd.c:
KNFised. `indent' worked 99% perfectly on bcd.c. It worked 99%
_imperfectly_ on subr_prf.c.
2) Use CURSIG(curproc) in PROCESS_ABORTING instead of junk code was there.
3) Reanimate timeout code in DO_SLEEP by setting WK_TIMEOUT flag
which is never set in old code.
4) DO_SLEEP: set aborting flag on interrupting singnals as supposed, not
on signals which do nothing as in old code.
5) Cleanup WAKE_UP macro, WK_WAKEUP not used.
6) Remove wrong typecasts in sleep/wakeup code.
and suiword(). They are no longer referenced in the machine-independent
code (I think fuiword() and suiword() were only used by old versions of
ptrace(), and copyoutstr() by old versions of exec).
Added `const' where appropriate.
Changed u_int to size_t' where appropriate.
Named last arg of copystr() and copyinstr() better.
This random address can be matched (with some probability) with another
sleep addresses from other drivers, which can cause strange sleep/wakeup
sequence. Rewrite this ugly code to do the right thing.
Add more features to the one remaining to handle the job:
+ signed quantity.
# alternate format
- left padding
* read width as next arg.
n numeric in (argument specified) default radix.
Fix the DDB debugger to use these.
Use vprintf in debug routine in pcvt.
The warnings from gcc may become more wrong and intolerable because
of this.
Warning: I have not checked the entire source for unsupported or
changed constructs, but generally belive that there are only a few.
Suggested by: bde
the standard macro dbtob(). The non-B_PAGING case now works well enough
to run newfs on a 32GB virtual drive.
Fixed numerous bogus variable types and one overflowing multiplication
in the B_PAGING case of vnstrategy(). Swapping to virtual drives larger
than 2GB might work now.
This fixes PR943.
ffs/ffs_vfsops.c:
ffs_statfs() multiplied by (100 - minfree) as part of calculating the
minfree percentage (complemented in 100%), so with the standard minfree
of 8, it was broken for file systems of size >= 1TB/92 = 11GB. Use the
standard freespace() macro instead. This also fixes a rounding bug (the
"Avail" count was sometimes 1 too small).
ffs/* (not fixed):
The freespace() macro multiplies by minfree, so with the standard
minfree of 8, it is broken for file systems of size >= 1TB/8 = 128GB.
This bug is more serious since it affects block allocation.
ffs/ffs_alloc.c (not fixed):
Ordinary users are sometimes allowed to allocate 1 (partial) block
too many so that the "Avail" count goes negative. E.g., if there is
1 fragment available and the file is fairly large, one more full
block is allocated.
df/df.c:
ufs_df() used/uses essentially the same code as ffs_statfs(), so it
had/has the same bugs.
ufs_df() gratuitously replaced "Avail" counts of < 0 by 0, so it
gave different results for non-mounted file systems in this case.
SPIORDY just before we ack on the bus so that there is no chance to
see SPIORDY for the same byte twice.
Make some small modifications so that the Linux aic7xxx driver can use
our sequencer and register definition files verbatum.
(To be replaced by sysctl accesses some time ...)
Remove a backward jump from the NCR script, which allowed a SCSI target
to receive any number of NOP messages it desired. If a target indeed
does such a silly thing, make it fail at the next instruction, instead
of causing a timeout a few seconds later.
it out fixes my problem but hoses the GUS MAX probe messages. Check what
device we have and print things appropriately for each.
Pointed out by: Jim Lowe <james@miller.cs.uwm.edu>
local address, that was assigned with ifconfig alias and netmask
0xffffffff, would receive duplictae udp packets.
This behaviour can easily be seen by having named run, and using the alias
address as the name server.
This solution is not the pretiest one, but after talk with Garreth, it
is seen as the most easy one.
variants, idea taken from NetBSD clock.c.
At least year calculation was wrong, pointed by Bruce.
Use different strategy to store year for BIOS without RTC_CENTURY
