blocks of zeros could wind up in a file written to over NFS by a client.
The problem only occurs a few times per several gigabytes of data. This
problem turned out to be bug #3 below.
bug #1:
B_CLUSTEROK must be cleared when an NFS buffer is reverted from
stage 2 (ready for commit rpc) to stage 1 (ready for write).
Reversions can occur when a dirty NFS buffer is redirtied with new
data.
Otherwise the VFS/BIO system may end up thinking that a stage 1
NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable.
bug #2:
B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short
buffers only occur near the EOF of the file). Change to only set
when the buffer is a full biosize (usually 8K). This bug has no
effect but should be fixed in -current anyway. It need not be
backported.
bug #3:
B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is
typically only called by the update daemon). nfs_flush()
does a multi-pass loop but due to the lack of vnode locking it
is possible for new buffers to be added to the dirtyblkhd list
while a flush operation is going on. This may result in nfs_flush()
setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its
stage 1 write, causing only the commit rpc to be made and thus
causing the contents of the buffer to be thrown away (never sent to
the server).
The patch also contains some cleanup, which only applies to the commit
into -current.
Reviewed by: dg, julian
Originally Reported by: Dan Nelson <dnelson@emsphone.com>
means that running out of mbuf space isn't a panic anymore, and code
which runs out of network memory will sleep to wait for it.
Submitted by: Bosko Milekic <bmilekic@dsuper.net>
Reviewed by: green, wollman
madvise().
This feature prevents the update daemon from gratuitously flushing
dirty pages associated with a mapped file-backed region of memory. The
system pager will still page the memory as necessary and the VM system
will still be fully coherent with the filesystem. Modifications made
by other means to the same area of memory, for example by write(), are
unaffected. The feature works on a page-granularity basis.
MAP_NOSYNC allows one to use mmap() to share memory between processes
without incuring any significant filesystem overhead, putting it in
the same performance category as SysV Shared memory and anonymous memory.
Reviewed by: julian, alc, dg
add support for non-pnp cards to sbc
move card identification to sbc
channel-swapping code is in sb now instead of dsp
vibra16x support is still broken, but will be fixed soon
note: sbc is now compulsory for sb cards
for pnp cards use:
device sbc0
for non-pnp cards eg:
device sbc0 at isa? port 0x240 irq 5 drq 3 flags 0x15
(hints as oldpcm)
both in addition to:
device pcm0
Reviewed by: tanimura,dfr
Said he liked it: peter
Currently we have a problem in that `dhclient' bails when configuring the
second interface as port 68 is already in use (by the `dhclient' started
for the first interface).
PR: 14810
Submitted by: n_hibma
Doing so is bogus if the loop-back interface was not configured.
Typically ``network_interfaces="auto"'' will return the list of
interfaces such that "lo0" is not first. Thus there are times when
`dhclient' configs an interface before "lo0" is configured.
Under BSD4.4, there's no need to add the above route, as it will be
automatically generated by the kernel.
PR: conf/14098, misc/15183
Submitted by: luoqi (true identification of the problem)
second LUN to show up.
mjacob's change (which is correct) in rev 1.21 of cam_periph.c to elminiate
infinite retries of the SCSI busy status bit seems to have broken probing
of Pioneer changers that aren't already quirked.
The right way to fix this is probably to change things around so we can
guarantee 100% sequential probing of LUN-based changers even if they aren't
quirked. This should fix things for now, though.
* lockstatus() and VOP_ISLOCKED() gets a new process argument and a new
return value: LK_EXCLOTHER, when the lock is held exclusively by another
process.
* The ASSERT_VOP_(UN)LOCKED family is extended to use what this gives them
* Extend the vnode_if.src format to allow more exact specification than
locked/unlocked.
This commit should not do any semantic changes unless you are using
DEBUG_VFS_LOCKS.
Discussed with: grog, mch, peter, phk
Reviewed by: peter
is good for... :-)), I discovered that part of the change when mkstemps()
was brought in was missed - it was missing the termination case to make
sure it doesn't walk into the suffix. This isn't the same code OpenBSD
has, I think this is a little better as we terminate the loop in a better
spot.
last time.
There's a possibility that there are still things wrong with the
commit. I wasn't able to build the version I committed, and I won't
be for another few days. If any committer finds further omissions,
the original source archive is on freefall:~grog/picobsd.tar.gz. Feel
free to correct.
Reported-by: "Louis A. Mamakos" <louie@TransSys.COM>
The UPAGES have not been there since Jan '96, but the hole was preserved
for BSD/OS binary compatability. This has been fixed other ways (%ebx
now has a pointer to PS_STRINGS), and the stack is nowhere near where
it used to be so this hack isn't required anymore.
maps onto the upages. We used to use this extensively, particularly
for ps and gdb. Both of these have been "fixed". ps gets the p_stats
via eproc along with all the other stats, and gdb uses the regs, fpregs
etc files.
Once apon a time the UPAGES were mapped here, but that changed back
in January '96. This essentially kills my revisions 1.16 and 1.17.
The 2-page "hole" above the stack can be reclaimed now.
by the Linux emulator (and other emulators) for syscall argument
translation. The x86 port currently seems to allow unrestricted kernel
accesses to user memory.
Reviewed by: alc, gallatin
and fixing some major bugs.
- Add support for the v5 firmware interface, used by the DAC1164P
(tested) and AcceleRAID 352 (untested but should work). We now cover
all of the Mylex family's protocols except for v2 (used by EISA and
Alpha-compatible cards).
- Fix an accounting bug which resulted in endless 'poll still busy'
messages. In situations of high controller load the count of poll
commands could be incremented without actually successfully launching
a command. This totally removes the accounting for status poll
commnads; it was its own worst enemy.
- Add some simple reentry prevention locks to processing of the waiting
and completed command queues to prevent races which could result in
I/O being done or completed twice (both are fatal). This highlights
a need for simple locking primitives in both the UP and SMP kernels.
- Streamline the handling of command completion to reduce the amount of
redundant work being done. Remove the code which tests for commands
that have gone missing in action; nobody has ever seen one of these
and it wouldn't have worked properly anyhow.
- Handle disconnection of drives from the controller in the detach,
not shutdown method. This avoids problems flushing the cache in
a panic when a drive is mounted.
- Don't call bus_generic_detach when disconnecting drives; it doesn't
actually do anything useful.
- Increment the log message index regardless of whether we actually
retrieved one or not. If we run into a message that we can't fetch,
we don't want to spin endlessly complaining about the fact.
- Don't assume that interrupts will work when we're flushing the
controller. We may think they are enabled, but in eg. a panic
situation the controller may not be able to deliver an interrupt.
using make instead of custom scripts) and two floppies instead of
one. The resultant floppy can do everything that the individual
floppies (dial, net, install, isp, router) could do, modulo some bit
rot that has occurred since PicoBSD last compiled. It also includes
all the programs on the fixit floppy, which could thus also die.
/bin currently contains the following files:
-sh dump ln ns sps
[ ed login ping stty
badsect ex ls ps swapon
cat expr mkdir pwd sync
chgrp fdisk mknod pwd_mkdb sysctl
chmod find more rdump syslogd
chown fsck mount reboot tar
chroot ftp mount_cd9660 restore telnet
clri getty mount_msdos rlogin telnetd
cp grep mount_nfs rm test
date gunzip mount_std rmdir traceroute
dd gzip msg route umount
dev_mkdb hostname mt routed vi
df ifconfig mv rrestore view
dhclient inetd natd rsh vm
dhclient-script init netstat sed w
disklabel kget newfs sh zcat
dmesg kill nfs sleep
Structure is in place for using the same build for the other
directories, but I'm no longer sure we need this. The current first
floppy will run fine by itself, but the size of a compressed kernel
has increased by nearly 50% since 3.2, and there's not much space for
anything useful on the remainder of the floppy. The current method
creates a larger mfs and can read as many floppies as the user can
stand. The footprint appears to be round 14 MB.
using make instead of custom scripts) and two floppies instead of
one. The resultant floppy can do everything that the individual
floppies (dial, net, install, isp, router) could do, modulo some bit
rot that has occurred since PicoBSD last compiled. It also includes
all the programs on the fixit floppy, which could thus also die.
/bin currently contains the following files:
-sh dump ln ns sps
[ ed login ping stty
badsect ex ls ps swapon
cat expr mkdir pwd sync
chgrp fdisk mknod pwd_mkdb sysctl
chmod find more rdump syslogd
chown fsck mount reboot tar
chroot ftp mount_cd9660 restore telnet
clri getty mount_msdos rlogin telnetd
cp grep mount_nfs rm test
date gunzip mount_std rmdir traceroute
dd gzip msg route umount
dev_mkdb hostname mt routed vi
df ifconfig mv rrestore view
dhclient inetd natd rsh vm
dhclient-script init netstat sed w
disklabel kget newfs sh zcat
dmesg kill nfs sleep
Structure is in place for using the same build for the other
directories, but I'm no longer sure we need this. The current first
floppy will run fine by itself, but the size of a compressed kernel
has increased by nearly 50% since 3.2, and there's not much space for
anything useful on the remainder of the floppy. The current method
creates a larger mfs and can read as many floppies as the user can
stand. The footprint appears to be round 14 MB.
Work-sponsored-by: Sitara Networks Inc.
