done when the command-in-progress bit has been cleared. The 1ms
time that is mentioned also seems wrong; 20,000us is not enough
on at least one machine. So, do a delay(100000) in ex_reset.
Fortunately, this function is only called when the card is
attached or when hardware failure is suspected.
While here, only mask one 1 bit in the 'reset hack' case.
This is based on amiga's siop driver, but converted to use
bus_space(9) functions and modified to fit bus_dma(9) framework.
Currently tested on NetBSD/arc with jazzio 53c710 SCSI,
which really requires bus_dma(9) functions :-)
Sync transfers and disconnect/reconnect are also working.
TODO:
- Test under more heavy load
- Clean up osiop_checkintr() hander
- Reorganize command queue and sync negotiation handling more suitable
for thorpej-scsipi mid-layer
- Re-think defered interrupt handling for amiga
reasonable transmit/receive buffer count.
This is needed for e.g. SKNET adapters, which use top 30 bytes of 16KB
memory to map registers and PROM and hence not all the memory is
available for buffers.
to 16 or 4 (depending on capabilities of adapter), as it was before
thorpej_scsipi integration
Waiting feedback to known whenever the problem with openings set to AHC_SCB_MAX
existed before.
- Add (missed)powerhook_disestablish() in ex_detach().
- Sync with below. Original commit log message:
Add new powerhook argument values, PWR_SOFTSUSPEND, PWR_SOFTSTANDBY and
PWR_SOFTRESUME. Apm calls powerhook with the values in normal interrupt
priority level while others are protected with splhigh().
This is a completely rewritten scsipi_xfer execution engine, and the
associated changes to HBA drivers. Overview of changes & features:
- All xfers are queued in the mid-layer, rather than doing so in an
ad-hoc fashion in individual adapter drivers.
- Adapter/channel resource management in the mid-layer, avoids even trying
to start running an xfer if the adapter/channel doesn't have the resources.
- Better communication between the mid-layer and the adapters.
- Asynchronous event notification mechanism from adapter to mid-layer and
peripherals.
- Better peripheral queue management: freeze/thaw, sorted requeueing during
recovery, etc.
- Clean separation of peripherals, adapters, and adapter channels (no more
scsipi_link).
- Kernel thread for each scsipi_channel makes error recovery much easier
(no more dealing with interrupt context when recovering from an error).
- Mid-layer support for tagged queueing: commands can have the tag type
set explicitly, tag IDs are allocated in the mid-layer (thus eliminating
the need to use buggy tag ID allocation schemes in many adapter drivers).
- support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command
will be requeued, or a REQUEST SENSE will be sent as appropriate.
Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
a happy code sweep (in the kernel!) and constified a random selection of
kernel objects. This changed the alignment of the previously-aligned-by-
accident seqprog array, and exposed a lurking bug. I can't decide if this
is good or bad.
saves about 2.2MB under /usr/include/dev/. Discussed on tech-kern@
recently.
I HOPE to get the list right. The headers I left in are ones
used for MI tools and those whose usage I discovered by grep over tree sources.
Feel free to put needed includes back in if you encounter anything which
should not be removed from lists.
mouse port seem to interpret the "test aux port" (0xa9) command differently,
leading to a non-working keyboard.
Now we try to echo a byte through the aux port by means of the "echo aux"
(0xd3) command, which is what Linux does.
Thanks to Christoph Badura for detailed reports and testing.
of system drives from an ENQUIRY.
- Complain if there's more than 1 segment when issuing some kind of enquiry
(this needs to be fixed properly).
- Fix the ID hack for 2.xx firmware.
- Fix an argument to bus_dmamap_sync().
xx,f0. This appears to heavily alleviate, but not to eliminate entirely,
the problems I've been seeing with garbage being read from the rx buffer.
I suspect the real solution lies elsewhere.
* Use the timer to timoe out transmit operations.
* Spot when the "next packet" pointer falls outside the recieve buffer and
reset the interface.
* Don't reset the interface when we get a bad packet (unless there's
something else wrong as well).
About the only bit of his code not here is the transmit routines, which I'll
merge in separately.
Also a few bug-fixes, so (for instance) multicast on an 8005 doesn't
immediately fall back to IFF_ALLMULTI.
This now provides slightly more functionality than the FreeBSD layer1-newbus
interface. It was meant to be a simple change to one header and a few
c files, but the change rippled all through various stuff.
To prevent a change to the kernel<->userland interface right now the kernel
is now lying about card types to userland (but who cares). This will be fixed
when the userland interface changes, after layer 3 <-> layer 4 has been
fixed.
Functional changes:
Provide a clean interface for hardware drivers to attach to the upper
layers. This will need another small change in the B-channel handling
when a similar change to the layer 3 <-> layer 4 interface happens.
Avoid passing indices into global arrays of pointers around, instead pass
the pointers itself. Don't code hardware driver types by predefined magic
numbers (think LKM). Prepare for detachable drivers (think pcmcia).
While there remove some sets of function pointers always pointing to the
same function (meant to be the configurable set of D channel protocol
handlers). It is unlikely another supported D-channel protocol will fit into
that (maximal layer interface) abstraction. When we get support for another
protocol, we will need to come up with a workable interface. Besides, the
old implementation was, uhm, strange.
provided to me by David Brownlee (thanks!).
Performance of this card is quite poor on my PS/2 with 386DX, like 100KB/s
at best, but as low as 5KB/s when transferring bigger files due to
packet overruns. It would be good to revisit this later, probably by
teaching the ic code to use RX Early.
up header guard. Clean up structure definition for at2_entry so
it's correct rather than derived from at_entry- this is important
for unswizzling purposes. Add a whole bunch of unswizzle macros-
they're not quite right yet but at least they're a start. Note that
we now have, for at_entry, a 16 bit firmware handle as part of what
had been at_reserved- this is to correlate ATIOs with CTIOs- and
this must be carried along as part of a tag value to use with all
CTIOs we send in relation to this ATIO. ANSIfy.
- Load the DMA address a byte at a time, and
- Use bus_space_read_stream_2() instead of bus_space_read_2() when reading
the byte-stream of configuration data from the card.
Sanity-checked on both i386 and sandpoint. Reviewed by ad.
for the ICB firmware options meant- *I* had taken it to
mean that if you set it, Node Name would be ignored and
derived from Port Name. Actually, it meant the opposite.
As a consequence- change ICBOPT_USE_PORTNAME to the
define ICBOPT_BOTH_WWNS- makes more sense.
isp_iid_set/isp_iid for fibre channel- this is because we now
fake a port database entry for ourselves. Add the additional loop
states between LOOP_PDB_RCVD and LOOP_READY.
Change and comment on a wad of Fibre Channel isp_control functions.
Change and comment on some of the ISPASYNC Fibre Channel events.
Add was_fabric_dev/fabric_dev tags to our local FC database structure
(so we can see rapidly whether something was a fabric device but is
now gone).
Add a tag which says what role this adapter should take. It can take
on the value of None, Target, Initiator or Both. None is useful for
warm failover purposes. Remove the ISP_CFG_NOINIT silliness since
a role of "None" does this.
Add a isp_lastmbxcmd tag to store the opcode for the last mailbox
command used.
to report the mailbox command that times out. Fix isp_unswizzle_sns_rsp
which for reasons *I* find obscurer just doesn't work correctly on sparc64
with words past 128. I have no idea why this *does* work on SparcLinux.
and depending on role, make sure link is up, scan the fabric (if we're
connected to a fabric), scan the local loop (if appropriate), merge
the results into the local port database then, check once again
to make sure we have f/w at FW_READY state and the the loopstate
is LOOP_READY.
Remove ISP2100_FABRIC defines- we always handle fabric now. Insert
isp_getmap helper function (for getting Loop Position map). Make
sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE
for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic
f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP,
and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications
upstream for changes in the name server, changes in the port database, and
f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event.
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol
Module) and FBM (Fibre Buffer Modules). Also remember to clear the
semaphore registers. Tell the RISC processor to not halt on FPM
parity errors.
Finally, use a new tag in the softc to store the opcode for the
last mailbox command used so we can report which opcode timed
out.
Major stuff:
Quite massively redo how we handle Loop events- we've now added several
intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us
a lot finer control about how we scan fabric, whether we go further
than scanning fabric, how we look at the local loop, and whether we
merge entries at the level or not. This is the next to last step for
moving managing loop state out of the core module entirely (whereupon
loop && fabric events will simply freeze the command queue and a thread
will run to figure out what's changed and *it* will re-enable the queu).
This fine amount of control also gets us closer to having an external
policy engine decide which fabric devices we really want to log into.
Throw out the ISP_CFG_NOINIT silliness and instead go to the use of
adapter 'roles' to see whether one completes initialization or not
(mostly for Fibre Channel). The ultimate intent, btw, of all of this
is to have a warm standby adapter for failover reasons. Because
we do roles now, setting of Target Capable Class 3 service parameters
in the ICB for the 2x00 cards reflects from role. Also, in isp_start,
if we're not supporting an initiator role, we bounce outgoing commands
with a Selection Timeout error. Also clean out the TOGGLE_TMODE
goop for FC- there is no toggling of target mode like there is
for parallel SCSI cards.
Do more cleanup with respect to using target ids 0..125 in F-port
topologies. Also keep track of things which *were* fabric devices
so that when you rescan the fabric you can notify the outer layers
when fabric devices go away.
Only force a LOGOUT for fabric devices if they're still logged in
(i.e., you cat their Port Database entry. Clean up the Get All Next
scanning.
- run request sense command without disconnect
- don't restart the script before siop_scsicmd_end has been called if the
cmd didn't complete with good status.
- reserve slot 0 for request sense, to make sure it'll be sent first.
transmit jobs. The previous logic occasionally, under heavy
load, would attempt to sync and unload DMA maps that weren't
currently in use, and also occasionally leaked mbufs under
heavy load.
Yay, NFS now works pretty well on my Dreamcast.
it causes assertions to fail in the PCI bounce buffer code. On
other platforms, it causes the address of the next packet to
be mis-calculated, leading to packet loss.
if we are requested range of multicast address or too many multicast address,
program multicast filter to receive all multicast address. And set/clear
IFF_ALLMULTI flag properly.
This adds support for EtherExpress/16 cards with 16k of RAM, and in the
process adds general support for PIO mode on these cards. This entails
changing the way the i82586 driver handles bus barriers, since it doesn't
allow for strange cases like this.
This has been tested on the i386 port with the 'ix' driver in both
16KB (which was the source of the problem) and 32KB modes, as well
as with the 'ef' driver. I've tested it (briefly) with 'ei' on arm26
as well. In theory, drivers other than 'ix' should follow precisely the
same code paths as before.
the link level name for the interface (ifp->if_sadl) is allocated
before ifp->if_addrlen is initialized, which could lead to allocating
too little space for the link level address.
Do this by splitting allocation of the link level name out of
if_attach() and into if_alloc_sadl(), which is normally called
by functions like ether_ifattach(). Network interfaces which
don't have a link-specific attach routine must call if_alloc_sadl()
themselves (example: gif).
Link level names are freed by if_free_sadl(), which can be called
from e.g. ether_ifdetach(). Drivers never need call if_free_sadl()
themselves as if_detach() will do it if it is not already done.
While here, add the ability to pass an AF_LINK address to
SIOCSIFADDR in ether_ioctl() (this is what caused me to notice
the problem that the above fixes).
and place the definition in <machine/types.h>. This can now be used
as a flag to indicate whether or not <machine/intr.h> can be included
to get the generic soft interrupt API.
This means we should be able to work with McData switches now. Change
ISPASYNC_PDB_CHANGED to ISPASYNC_LOGGED_INOUT (more descriptive). Allow
F-Port topologies to use target ids 0..125 to log into fabric devices.
Yet again fool around with defaul WWN stuff.
Rumors say there are archs without ISA busses, so avoid including
(uneccesarily) isa bus headers in MI files.
XXX this is the minimal solution, layer interface calls will have
XXX to be revisited later
be attached with this flag.
Some CF Card (for ex. IBM MicroDrive and SanDisk) doesn't seem to implement
drive select command. In this case, you can't eliminate ghost drive properly.
So you should use this flag to ignore the ghost by force.
It's probably not really a compiler bug- somebody pointed out
that it was the kernel making strings readonly. But I do think it's
a bug. The actual code was really more like:
char *revname;
...
revname = "2X00";
...
revname[1] = '2'; <<<<<<<<< BOOOM
The variable revname is not a const. If I had said
const char *revname = "2X00"
...
revname[1] = '2';
that would indeed be breaking const rules.
char *foo = "XXXX";
...
foo[1] = 'Y';
blow up (in the kernel) with the 2nd assignment. Work around it here-
it's probably just as well- I was spending more in cpu instructions doing
the assignment than I was saving in string space (it would have been
cheap on a pdp11 or a 68k- but the address loads and assignments on something
like sparc or alpha way outweigh the savings in space. Tsk.).
change I didn't take is the %llu format- I can't have a common
across multiple platform module assume a %ll argument capability-
which really pointed out that I shouldn't be trying to *print*
something which could long long.
process steps (this is used in some platforms where you want to bring
the adapter to ready (to get a WWN, e.g.) , but not engage either target
or initiator mode until some later time). Set the correct defaults WWNs.
