revision 1.21
date: 1999/10/15 17:29:20; author: imp; state: Exp; lines: +3 -3
Reorganize the attachement point for pcic (it was unattached and
floating before). Attach pccard devices to pcic, one per slot
(although this may change to one per pcic). pcic is now attached to
isa (to act as a bridge) and pccard is attached to pcic, cbb and
pc98ic (the last two are card bus bridge and the pc98ic version of
pcic, neither of which are in the tree yet). Move pccard compat code
into pccard/pccard_compat.c.
THIS REQUIRES A CONFIG FILE CHANGE. You must change your pcic/card
entries to be:
# PCCARD (PCMCIA) support
controller pcic0 at isa?
controller pcic1 at isa?
controller card0
The old system was upside down and this corrects that problem. It
will make it easier to add support for YENTA pccard/card bus bridges.
Much more cleanup needs to happen before newbus devices can have
pccard attachments. My previous commit's comments were premature.
Forgotten by: imp
- Fix a bug in rl_rxeof() handler: in the case where the packet wraps
from the end of the receive buffer back to the beginning, we need to
insure that at least sizeof(ether_header) bytes make it into the first
mbuf. If we don't, then doing eh = mtod(m, struct ether_header *)
loses. To avoid this, we use m_pullup() to suck at least MHLEN -
RL_ETHER_ALIGN bytes into the first mbuf, which should also help
small packets fit into a single mbuf.
Pointed out by: Philip A. Prindeville <philipp@zembu.com>
- Make the transmit threshold autotuning: start off with a small value
and jack it up when TX underruns are detected.
- Also improve TX error recovery: kick the chip in the head with a
reset/init sequence to make sure it recovers afer a transmit error.
glibc2 defines struct dirent differently than the Linux kernel does.
The getdents function therefore needs to read a heuristically defined
number of kernel dirents to satisfy the request. In case where too
many kernel dirents have been read, the function lseeks on the
directory so that a next call will start with the right dirent. The
offset used in lseeking is the offset-field in the last dirent passed
to the application. This can only mean that the offset-field holds
the offset of the next dirent and not the offset of the dirent itself.
Been in production for 3 years now. Gives Instant Frame relay to if_sr
and if_ar drivers, and PPPOE support soon. See:
ftp://ftp.whistle.com/pub/archie/netgraph/index.html
for on-line manual pages.
Reviewed by: Doug Rabson (dfr@freebsd.org)
Obtained from: Whistle CVS tree
to remove 'b'lock devices. The agreement is, essentially, that
block devices will be collapsed into character devices as a first
step (though I don't particularly agree), and raw device names 'rxxx'
will become simply 'xxx' in devfs in the second step (i.e. no 'rxxx'
names will exist). The renaming will not effect the original /dev
and the expectation is that devfs will eventually (but not immediately)
become the standard way to access devices in the system.
If it is determined that a reimplementation of block device access
characteristics is beneficial, a number of alternatives will
be possible that do not involve resurrecting the 'b'lock device class.
For example, an ioctl() that might be made on an open character device
descriptor or a generic buffered overlay device.
This commit removes the blockdev disablement sysctl which does not
apply to the solution that was reached.
and unusable by the pccard system since pccard doesn't attach to the
nexus any more. This was stopping my 3c589D from working as pccard unit
0 is used directly for resource allocation and this fails when unit 0
isn't actually attached to anything.
it has the same value on all platforms. Previously it was just under
3 seconds on x86 (typically hz<=128) and just under 1/3 of a second on
alpha (typically hz>=1024). This covers up a race between ad_interrupt()
and ad_timeout() which is being looked into.
reviewd by: sos
This means that access to block devices nodes will act the
same as char device nodes for disk-like devices.
If you encounter problems after this, where programs accessing
disks directly fail to operate, please use the following command
to revert to previous behaviour:
sysctl -w vfs.bdev_buffered=1
And verify that this was indeed the cause of your trouble.
See the mail-archives of the arch@FreeBSD.org list for background.
bootup. Somehow my backout of an abortaive attempt at shared
memory autoconfiguration included this line:
sc->mem_shared = 1;
Which is fairly important as it turns out.
Since I performed my pre-commit testing on a different box with a generic
NE2000 I didn't catch this. Pointy hat.
Put splbio protection around the main launch loop. We've seen cases where
the bottom half was cutting off the branch on which we're sitting.
Experienced-by: Michael Reifenberger <root@nihil.plaut.de>
have you is prototyped). Removed code versions in md struct- not used
any more. Allocate transfer dma maps and xflist stuff in mbxdmasetup based
upon isp->isp_maxcmds. Allow for multiple calls to mbxdmasetup (for
isp_reset cases).
file later. Do some pencil-sharpening types of minor changes. Change
how active commands are remembered (using new inline functions to get
handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as
outgoing mailbox 2 will tell you the f/w's notion of the max commands
that can be supported. Attempt to retrieve loop topology. Add in the
appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little
Endian machines but is needed for sparc (on other platforms)). Move
the temp port database we use to find out where things have moved to
after a LIP to the softc and off the kernel stack. Follow Qlogic's
hint and don't bother setting a tag for commands that don't have
this enabled (presumably the f/w will do it's own selection then).
Use an INT_PENDING macro to check for an interrupt. The call to
ISP_DMAFREE now just takes the handle- not the 'handle-1' which was
a layering violation. Use CFGPRINTF in a couple of places to make
things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
where it defaults to one. Change simq width allocation to the max number
of commands supported by the HBA after f/w fires up- not the constant
MAXISPREQUEST value. Do some stylistic changes.
Add in null SWIZZLE definitions. Add in CFGPRINTF define. Change default
debug level to refer to an external isp_debug variable. Remove inline
functions as they're now in isp_inline.h and include that file.
the result queue length is never less than 64. Move (ick) temp port
database used for post-LIP merging off the kernel stack and put it
into the softc. Remove some target mode stuff which will come back
later in a different file. Change how the list of outstanding commands
are stored (now allocated at mailbox setup time to be just enough for
the max for a specific HBA which can vary). Keep a rotating seed of
the last index for this in the softc. Increase the count of active
commands from 10 to 16 bits.
phk's script walked through .c and .h files, but some of the ones on
the list are actually derived from sys/svr4/syscalls.master. Make
the necessary changes here and the others will implicitly follow...
Submitted by: phk
retry count for the ccb). This is probably not quite the right thing, but it
is better than silently hanging on (possibly broken) h/w which is what we
do now.
Reviewed by:Justin/Ken: they weren't entirely happy about it but didn't say no.
- Generally clean things up.
- PnP now supported.
Will convert to bus_space, ifmedia and add a DEVICE_IDENTIFY() method
for autodetection. As it stands
device ex0 at isa0
should find a card if one is present.
I feel less dirty now.
In order to make this work, I created a pseudo-PHY driver to deal with
Macronix chips that use the built-in NWAY support and symbol mode port.
This is actually all of them, with the exception of the original MX98713
which presents its NWAY support via the MII serial interface.
The mxphy driver actually manipulates the controller registers directly
rather than using the miibus_readreg()/miibus_writereg() bus interface
since there are no MII registers to read. The mx driver itself pretends
that the NWAY interface is a PHY locayed at MII address 31 for the sole
purpose of allowing the mxphy_probe() routine to know when it needs to
attach to a host controller.
pc98 case that I missed before. Attempt to get the irq for the PCIC
first from the loader env var and second from the config system. I've
been able to boot my laptop with a kernel that hardwired the irq to
10. This should allow boot -c to finally start working for pcic irq,
but I've not tested that. Add $FreeBSD$ to slot.h.
32 bytes is safe.
Handle successful completion of message log retrieval commands.
With these changes, the driver correctly handles the consequences of drive
death and replacement in a reliable array. Note that the massive backlog
of I/O during handling of such an event can kill the system if softupdates
is enabled.
floating before). Attach pccard devices to pcic, one per slot
(although this may change to one per pcic). pcic is now attached to
isa (to act as a bridge) and pccard is attached to pcic, cbb and
pc98ic (the last two are card bus bridge and the pc98ic version of
pcic, neither of which are in the tree yet). Move pccard compat code
into pccard/pccard_compat.c.
THIS REQUIRES A CONFIG FILE CHANGE. You must change your pcic/card
entries to be:
# PCCARD (PCMCIA) support
controller pcic0 at isa?
controller pcic1 at isa?
controller card0
The old system was upside down and this corrects that problem. It
will make it easier to add support for YENTA pccard/card bus bridges.
Much more cleanup needs to happen before newbus devices can have
pccard attachments. My previous commit's comments were premature.
Without it the kernel config options, like OVERRIDE_TUNER, where not
getting passed to the driver.
Bug noticed by: Marc Fonvieille <fonvi@club-internet.fr>
completion' flag. If set, the interface output routine will assume that
the packet already has a valid link-level source address. This defaults
to off (the address is overwritten)
PR: kern/10680
Submitted by: "Christopher N . Harrell" <cnh@mindspring.net>
Obtained from: NetBSD
now lives in the respective bus front end files.
- Add various function prototypes to if_edvar.h
- Clean up some debugging code that snuck into if_ed_isa.c
- Turn on the right bits in files.i386
devices. There may still be problems with said drivers, if so please
let me know.
o Move attach-like functionality to the nbk attach compatibility code.
o Smarter probe code: for the compatibility code probe succeeds if
strcmp succeeds, for noncompatibility you can do anything you like.
o Get rid of some compiler warnings introduced in last commit.
- When setting/clearing promisc mode, just update the filter, don't
reset the whole interface.
- Call xl_init() in xl_ifmedia_upd() when setting miibus media modes. This
fixes a problem with the 3c905B-COMBO where switching from 10base5/AUI
or 10base2/BNC to a 10/100 mode doesn't always work right.
- Attempt to reset the interface in xl_init() so that we know we're getting
the receive and transmit rings reset properly.
* Change the hack used on the alpha for mapping devices into DENSE or
BWX memory spaces to a simpler one. Its still a hack and should be
a seperate api to explicitly map the resource.
* Add $FreeBSD$ as necessary.
* Move pnp_eisaformat() to pnp.c, declared in <isa/pnpvar.h>.
* Turn the pnpbios code into an enumerator for the isa bus. This allows
all devices known to the bios to be probed automatically.
Currently the pnpbios code is dependant on the PNPBIOS option. As the code
is tested more and when more drivers are converted this will be made the
default. I have PnP changes in the wings for fdc, atkbd, psm, pcaudio, and
joy. Sio already works with pnpbios.
PJ cards. This will probably also support the AcceleRAID and
eXtremeRAID cards, but nobody has volunteered one for testing, so I
haven't enabled their PCI device IDs.
Slightly clean up communication between the disk devices and the
controller device as per new practice, and move some more register-
related items int mlxreg.h from mlxvar.h.
Remove some unnecessary read-modify-write operations to the card
control registers; they don't behave like that.
Increase the status polling interval to 10 seconds. It's still possible
to load the card up to the point where a status poll will find the
previous poll still running, but this will reduce the incidence of
complaints.
maximum of 64 kB.
vinum_conf, struct drive: add fields for the current number of active
requests and the maximum ever active.
struct sd: Add fields for initialize progress.
struct mc: set the size of saved file names to MCFILENAMELEN instead
of the previous explicit constant.
limit the number of outstanding requests on a specific drive and
overall.
Change the way we set the active request count. This enables us to
start the requests without being in splbio for the duration, which
could be very long for IDE drives in PIO mode.
context. Be prepared to fail instead.
MMalloc, FFree: ensure that saved file names are properly terminated.
Use MCFILENAMELEN instead of the previous explicit constant.
and then doing it itself, resulting in a panic downed drives.
Sleuth-work-by: Christopher Masto <chris@netmonger.net>
Remove dummy function initsd, which is now (implemented) in
vinumrevive.c
vinum_scandisk: Check that a drive is up before reading from it. This
is probably excessive paranoia.
same interface as Intel's P6 family has. Incidentally, I had disabled
it in the first place since I knew the K7s were coming out soon but
did not want to assume they'd have the same MTRR interface as Intel's
chips.
Submitted by: Ville-Pertti Keinonen <will@iki.fi>
