make ntfs_ntref() real function and properly lock ip->i_interlock when
updating ip->i_usecount
g/c ntfs_uustricmp()
make ntfs_uastr[i]cmp() static, rename arguments to be a bit more mnemonic
add couple of casts to NTFS_TOUPPER() and ntfs_uastr[i]cmp() - this
corrects couple of cases when open()/stat() failed for
files which apparently existed and even showed in ls output
ntfs_ntlookup(): initialize ip properly if ntnode is found in hash table,
so that debug prints doesn't print garbage in that case
some whitespace policy
With nfsv2, the nfsm_reply() macro always causes the service routine
to return if error was nonzero.
With nfsv3, we can keep going after nfsm_reply() without returning,
but nqnfsrv_getlease() didn't take this into account, so add a
return(0) after each error-case nfsm_reply(0).
stack-allocate a sockaddr_storage for the temporary sockaddr rather
than putting it in an mbuf.
neighbor discovery wants to delete expired ifa's from a timeout
handler. allowing rtinit(RTM_DELETE, ...) to run at interrupt level
allows this to work.
i think we can afford the extra ~128 bytes of stack depth ..
shared by several fnodes when the file is simultaneonsly opened by
several callers and different "attributes" (a.k.a. resource forks)
are requested. They have been converted to use lockmgr()-style locking.
guard the setting and checking of i_usecount by simplelock; as
situation where there would be several users of same ntnode is
extremery rare, it's not worth the efford to change the code to
hold the (exclusive) lock only when it's needed
* in struct fnode, g/c f_dev & f_type (former is already in ntnode,
latter is in parent vnode), move f_devvp from struct fnode into
struct ntnode (no reason this should be copied in each fnode)
do a gross hack which allows seemingly-broken quantum drives to function
with this driver. The gross hack is to disable tagged queueing completely
when QUEUE FULL is received. That costs performance on drives which
do tagged queueing properly and which return QUEUE FULL, but given the way
this driver works it's seems to be the only thing short of significant
recoding which will make it function with the quantum drives in question.
2. pull in a fix from FreeBSD:
revision 1.18
date: 1999/09/17 18:04:14; author: wpaul; state: Exp; lines: +3 -3
Remember to account for ETHER_ALIGN when setting the maxmimum packet
length for mini receive ring. The max length was MHLEN, however the mbufs
are actually shortened to MHLEN - ETHER_ALIGN to force payload alignment.
queueing support that decreases the number of openings on a device; it
previously assumed that a scsipi_link's `openings' were descreased as
commands were issued, which is not longer the case (`active' is increased).
1. Add power control to reduce power consumption.
The behavior can be controlled by mixerctl(1).
2. Add suspend/resume hook to keep the parameters.
Currently this includes CODEC and control registers but
does not include the DMA state. FIXME.
3. Support "opl at ym" and "mpu at ym" attachment.
Actually, this is required for power control.
4. Add support for the on-chip 3D enhancement hardware.
This includes bass/treble enhancement, wide stereo, and
the equalization mode. This can be controlled by mixerctl(1).
5. Make "audioctl -w play.gain=xxx" work like as
"mixerctl -w inputs.dac=xxx" and now it changes
the wave output volume correctly.
6. Support hardware master volume switches.
7. Change formula of calculating gain so that all the levels
should be used equally.
8. Change "audioctl name" to "OPL3-SA3" and "audioctl version"
to the version of the chip.
- Mute wave output, not the mixer output, on changing CODEC settings.
- Add support for muting wave output while playback or monitoring is inactive.
- Change formula of calculating gain so that all the levels should be
used equally (the old code uses the min/max gain only at 0/255).
- Cleanup
Implement this by making the fields `pm_reg_ptp' and `pm_reg_ptp_pa'
pointers to an array of `ncpu' region table pointers and corresponding
physical addresses.
This is a somewhat unfortunate side effect of having the per-CPU data
addressable by the same virtual address on each CPU.
- In dpt_intr(), if no interrupt is signalled but the HBA has indicated
that more data will be available soon (HA_ST_MORE), save a context switch
and wait for the data.
to the SCSIPI subsystem.
- Determine more characteristics about the board, including the number
of concurrent commands that the processor can execute at a time.
- Rework memory allocation strategy so that we can do dynamic pre-allocation
of resources according to workload.
Thanks to the Linux Buslogic driver for having some useful comments.
until Matt Jacob has a change to update the driver for the latest
firmware, etc. since update device parameters seems to fail once the
system is really up and running, and eventually causes the controller
to wedge. This may be due to a firmware bug.
Per discussion with Matt.
- `flags' is now gone, replaced with `xs_control' and `xs_status'.
- Massive cleanup of the control flags. Now we explicitly say that
a job is to complete asynchronously, rather than relying on side-effects,
and use a new flag to now that device discovery is being performed.
- Do SCSI device discovery interrupt-driven.
- 82C115 has a 128-bit multicast hash table, not 512-bit.
- Correct the way the MAC address is read from the SROM, after re-reading
the MX98715A Application Note.
Other semi-related changes:
- Differentiate between MX98715 and MX98715A.
- Improve the Macronix link-up/link-down detection.
this by frobbing the 10TCTL register minimally, toggling only the bit we
care about (and hoping that link-capability advertisement happens correctly).
MX98715 can talk to the network now.
- Add a comment about the Macronix MX98713 -- it's almost exactly like
a 21140A, and doens't have the Nway registers that the other Macronix
chips have.
- Bus specific front-end is now responsible for reading EATA configuration
- EATA configuration data is now saved in the softc
- Make synchronous writes work properly
- Don't ignore HBA timeouts in dpt_cmd()
- Prefix eata_cfg's members with `ec_', not `dc_'
* f_lock is only needed for FreeBSD (for their totally disgusting
and insane way of doing generic vnode locking)
* g/c f_mp, it was write-only
* some comments added
hannken