occur due to np->n_size potentially changing if nfs_getcacheblk()
blocks in nfs_write().
Second, under -current we must supply the proper bufsize when obtaining
buffers that straddle the EOF, but due to the fact that np->n_size can
change out from under us it is possible that we may specify the wrong
buffer size and wind up truncating dirty data written by another
process.
Both problems are solved by implementing nfs_rslock(), which allows us
to lock around sensitive buffer cache operations such as those that
occur when appending to a file.
It is believed that this race is responsible for causing dirtyoff/dirtyend
and (in stable) validoff/validend to exceed the buffer size. Therefore
we have now added a warning printf for the dirtyoff/end case in current.
However, we have introduced a new problem which we need to fix at some
point, and that is that soft or intr NFS mounts may become
uninterruptable from the point of view of process A which is stuck waiting
on rslock while process B is stuck doing the rpc. To unstick process A,
process B would have to be interrupted first.
Reviewed by: Alfred Perlstein <bright@wintelcom.net>
tsunami systems and the PCI bus-numbering system of FreeBSD. Eg, the former
allows for 2 PCI bus 2's (one each on hoses 0 and 1) while the latter
needs to give each PCI bus a unique monotonically increasing number.
It has been fairly well tested and correctly maps machines with a ppb on
hose 1 as well as machines with ppbs on both hoses.
DS10s remain untested, as I do not have a pci card with a ppb which will
pass POST in a tsunami.
This is a house of cards.
main component in the southbridge chip to determine which VIA chip
we are dealing with.
Try to enable DMA on generic controllers that say they has the
capability, instead of relying on the BIOS to have set it up.
The variables "m_mclalloc_wid" and "m_mballoc_wid" were not in the
proper place. They should have been in uipc_mbuf.c and have been global,
not in mbuf.h and local per each file that uses mbuf.h.
Sorta bug fix:
In mbuf.h, the definitions of various things for KERNEL and not
KERNEL cases were very screwy. This fixes all of that which I could
find.
- Add a flag DC_TX_INTR_ALWAYS which causes the transmit code to
request a TX done interrupt for every packet. The PNIC seems to need
this to insure that the sent TX buffers get reaped in a timely fashion.
- Try to unreset the SIA as soon as possible after resetting the whole
chip.
- Change dcphy to support either 10/100 or 10Mbps only NICs. The
built-in 21143 ethernet in Compaq Presario machines is 10Mbps only
and it doesn't work right if we try to advertise 100Mbps modes during
autoneg. When restricted to only 10mbps modes, it works fine.
Note that for now, I detect this condition by checking the PCI
subsystem ID on this NIC (which has a Compaq vendor/device ID).
Yes, I know that's what the SROM is supposed to be for. I'm deliberately
ignoring the SROM wherever possible. Sue me.
The latter two fixes allow if_dc to work correctly with the built-in
ethernet on certain Compaq Presario boxes. There are liable to be quite
a few people using these as their home systems who might want to try
FreeBSD; may as well be nice to them.
Now if anybody out there has an Alpha miata with 10Mbps ethernet and
can show me the output from pciconf -l on their system, I'd be grateful.
cannot unilaterally pass data to a client it can reduce the physical
disk transaction overhead by reading larger blocks. This results in
better pipelining of requests/responses over the network and an almost
100% increase in cpu efficiency on the server. On a 100BaseTX network
NFS read performance increases from 8.5 MBytes/sec to 10 MB/sec (maxed
out), and cpu efficiency increases from 72% idle to 80% idle on the server.
Reviewed by: Alfred Perlstein <bright@wintelcom.net>
frame types. Currently it supports only IPX protocol and doesn't
affect existing functionality when not loaded.
Reviewed by: Ollivier Robert <roberto@keltia.freenix.fr>
Add a missing DELAY(1) in ata_wait.
Change the info from ad_version, so the ATA version from the disk can
be used to quantify the DAM modes valid for this drive, ie be more
selective with turning DMA on on older disks that should not support it..
Fix the probe for BIOS enabled DMA in the generic case, master/slave
was reversed in the test.
Check the return for ata_command in all cases, and print warnings if
it fails.
Call ata_dmainit with all dmamodes off when falling back to PIO mode,
that should take care of both the Promise & HPT366 controllers not
being able to handle the fallback...
Cleanup the printf's in the drivers, use the prober device name (if
possible) instead of ataN-master/slave.
1. Data written beyond end of pipe buffer, causing kernel memory corruption.
- Check that space is still valid after obtaining the pipe lock.
- Defer the calculation of transfer size until the pipe
lock has been obtained.
- Update the pipe buffer pointers while holding the pipe lock.
2. Writes of size <= PIPE_BUF not always atomic.
- Allow an internal write to span two contiguous segments,
so writes of size <= PIPE_BUF can be kept atomic
when wrapping around from the end to the start of the
pipe buffer.
PR: 15235
Reviewed by: Matt Dillon <dillon@FreeBSD.org>
are configured, and/or associated with a parent device. If you
receive a frame for a VLAN that's not in the list, you walk off
the end of the list. Boom.
Submitted by: C. Stephen Gunn <csg@waterspout.com>
PR: 15291
vlan interfaces it manages. This prevents the interface from
actually sending or receiving data.
Submitted by: C. Stephen Gunn <csg@waterspout.com>
PR: 15290
to print out protocol specific pcb info.
A patch submitted by guido@gvr.org, and asmodai@wxs.nl also reported
the problem.
Thanks and sorry for your troubles.
Submitted by: guido@gvr.org
Reviewed by: shin
needed for ages, but keeps getting cut/pasted into new Makefiles.
(Once apon a time it was used to activate mount arguments in
<sys/mount.h>, but that was killed with extreme prejudice long ago)
we use. The .c half is statically compiled into the kernel. It's kinda
silly to generate a .h file on the fly that has inlines to call the
.c stuff when the .c code is fixed.
Also, zap the special treatment for VFS_KLD modules. This treatment
applies to lots of things, not just VFS's.
the kernel while the vnode_if.h header is a bunch of inlines to call the
code that is in the kernel. Generating the .h file on the fly is kinda
bogus because it has to match the one compiled into the kernel.
IMHO we should have kern/vnode_if.c and sys/vnode_if.h committed in the
tree but that's another battle.
