mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY
twice. I mean, this is crucial- give the IOC a chance to get
ready.
If mpt_reset is called to reinit things, and we succeed, make
sure to re-enable interrupts. This is what has mostly led to
system lockup after having to hard reset the chip. Also, if
we think that interrupts aren't function in mpt_cam_timeout,
for goodness sake, turn them on again.
In read_cfg_header, return distinguishing errnos so the caller
can decide what's an error. It's *not* an error to fail to
read a RAID page from a non-RAID capable device like the FC929X.
Some whitespace fixes (removing spaces from ends of lines).
not hang the system for 5 seconds. If a TMF doesn't complete within,
oh, say 500ms, that's enough.
Put in a printout to catch mpt_recover_commands being activated with
no commands.
*both* SAS and FC, not just SAS.
b) Don't tell the chip we want it to do FIFO signalling if we actually
don't set up the address where the FIFO signal is supposed to be written
(oops).
automatically both SATA and SAS drives. The async SAS event handling we catch
but ignore at present (so automagic attach/detach isn't hooked up yet).
Do 64 bit PCI support- we can now work on systems with > 4GB of memory.
Do large transfer support- we now can support up to reported chain depth, or
the length of our request area. We simply allocate additional request elements
when we would run out of room for chain lists.
Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms.
There were no RAID cards available for me to regression test.
The error recovery for this driver still is pretty bad.
of volumes that might need administrator attention through device
specific sysctl to simplify device monitoring.
Submitted by: Deomid Ryabkov <myself at rojer dot pp dot ru>
an allocation. This fixes the malloc 'use after free' panic on boot that
many were seeing. It doesn't solve the problem of the allocations being
cached and then written past their bounds later. That will take more work.
Submitted by: kan
holders. The license that was approved for my changes to this driver
originally came from LSI, but the changes to the driver core are not
owned by LSI.
MFC: 1 day
o Add timeout error recovery (from a thread context to avoid
the deferral of other critical interrupts).
o Properly recover commands across controller reset events.
o Update the driver to handle events and status codes that
have been added to the MPI spec since the driver was
originally written.
o Make the driver more modular to improve maintainability and
support dynamic "personality" registration (e.g. SCSI Initiator,
RAID, SAS, FC, etc).
o Shorten and simplify the common I/O path to improve driver
performance.
o Add RAID volume and RAID member state/settings reporting.
o Add periodic volume resynchronization status reporting.
o Add support for sysctl tunable resync rate, member write cache
enable, and volume transaction queue depth.
Sponsored by
----------------
Avid Technologies Inc:
SCSI error recovery, driver re-organization, update of MPI library
headers, portions of dynamic personality registration, and misc bug
fixes.
Wheel Open Technologies:
RAID event notification, RAID member pass-thru support, firmware
upload/download support, enhanced RAID resync speed, portions
of dynamic personality registration, and misc bug fixes.
Detailed Changes
================
mpt.c mpt_cam.c mpt_raid.c mpt_pci.c:
o Add support for personality modules. Each module exports
load, and unload module scope methods as well as probe, attach,
event, reset, shutdown, and detach per-device instance
methods
mpt.c mpt.h mpt_pci.c:
o The driver now associates a callback function (via an
index) with every transaction submitted to the controller.
This allows the main interrupt handler to absolve itself
of any knowledge of individual transaction/response types
by simply calling the callback function "registered" for
the transaction. We use a callback index instead of a
callback function pointer in each requests so we can
properly handle responses (e.g. event notifications)
that are not associated with a transaction. Personality
modules dynamically register their callbacks with the
driver core to receive the callback index to use for their
handlers.
o Move the interrupt handler into mpt.c. The ISR algorithm
is bus transport and OS independent and thus had no reason
to be in mpt_pci.c.
o Simplify configuration message reply handling by copying
reply frame data for the requester and storing completion
status in the original request structure.
o Add the mpt_complete_request_chain() helper method and use
it to implement reset handlers that must abort transactions.
o Keep track of all pending requests on the new
requests_pending_list in the softc.
o Add default handlers to mpt.c to handle generic event
notifications and controller reset activities. The event
handler code is largely the same as in the original driver.
The reset handler is new and terminates any pending transactions
with a status code indicating the controller needs to be
re-initialized.
o Add some endian support to the driver. A complete audit is
still required for this driver to have any hope of operating
in a big-endian environment.
o Use inttypes.h and __inline. Come closer to being style(9)
compliant.
o Remove extraneous use of typedefs.
o Convert request state from a strict enumeration to a series
of flags. This allows us to, for example, tag transactions
that have timed-out while retaining the state that the
transaction is still in-flight on the controller.
o Add mpt_wait_req() which allows a caller to poll or sleep
for the completion of a request. Use this to simplify
and factor code out from many initialization routines.
We also use this to sleep for task management request
completions in our CAM timeout handler.
mpt.c:
o Correct a bug in the event handler where request structures were
freed even if the request reply was marked as a continuation
reply. Continuation replies indicate that the controller still owns
the request and freeing these replies prematurely corrupted
controller state.
o Implement firmware upload and download. On controllers that do
not have dedicated NVRAM (as in the Sun v20/v40z), the firmware
image is downloaded to the controller by the system BIOS. This
image occupies precious controller RAM space until the host driver
fetches the image, reducing the number of concurrent I/Os the
controller can processes. The uploaded image is used to
re-program the controller during hard reset events since the
controller cannot fetch the firmware on its own. Implementing this
feature allows much higher queue depths when RAID volumes
are configured.
o Changed configuration page accessors to allow threads to sleep
rather than busy wait for completion.
o Removed hard coded data transfer sizes from configuration page
routines so that RAID configuration page processing is possible.
mpt_reg.h:
o Move controller register definitions into a separate file.
mpt.h:
o Re-arrange includes to allow inlined functions to be
defined in mpt.h.
o Add reply, event, and reset handler definitions.
o Add softc fields for handling timeout and controller
reset recovery.
mpt_cam.c:
o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality,
such as event handling, into mpt.c leaving only CAM SCSI
support here.
o Revamp completion handler to provide correct CAM status for
all currently defined SCSI MPI message result codes.
o Register event and reset handlers with the MPT core. Modify
the event handler to notify CAM of bus reset events. The
controller reset handler will abort any transactions that
have timed out. All other pending CAM transactions are
correctly aborted by the core driver's reset handler.
o Allocate a single request up front to perform task management
operations. This guarantees that we can always perform a
TMF operation even when the controller is saturated with other
operations. The single request also serves as a perfect
mechanism of guaranteeing that only a single TMF is in flight
at a time - something that is required according to the MPT
Fusion documentation.
o Add a helper function for issuing task management requests
to the controller. This is used to abort individual requests
or perform a bus reset.
o Modify the CAM XPT_BUS_RESET ccb handler to wait for and
properly handle the status of the bus reset task management
frame used to reset the bus. The previous code assumed that
the reset request would always succeed.
o Add timeout recovery support. When a timeout occurs, the
timed-out request is added to a queue to be processed by
our recovery thread and the thread is woken up. The recovery
thread processes timed-out command serially, attempting first
to abort them and then falling back to a bus reset if an
abort fails.
o Add calls to mpt_reset() to reset the controller if any
handshake command, bus reset attempt or abort attempt
fails due to a timeout.
o Export a secondary "bus" to CAM that exposes all volume drive
members as pass-thru devices, allowing CAM to perform proper
speed negotiation to hidden devices.
o Add a CAM async event handler tracking the AC_FOUND_DEVICE event.
Use this to trigger calls to set the per-volume queue depth once
the volume is fully registered with CAM. This is required to avoid
hitting firmware limits on volume queue depth. Exceeding the
limit causes the firmware to hang.
mpt_cam.h:
o Add several helper functions for interfacing to CAM and
performing timeout recovery.
mpt_pci.c:
o Disable interrupts on the controller before registering and
enabling interrupt delivery to the OS. Otherwise we risk
receiving interrupts before the driver is ready to receive
them.
o Make use of compatibility macros that allow the driver to
be compiled under 4.x and 5.x.
mpt_raid.c:
o Add a per-controller instance RAID thread to perform settings
changes and query status (minimizes CPU busy wait loops).
o Use a shutdown handler to disable "Member Write Cache Enable"
(MWCE) setting for RAID arrays set to enable MWCE During Rebuild.
o Change reply handler function signature to allow handlers to defer
the deletion of reply frames. Use this to allow the event reply
handler to queue up events that need to be acked if no resources
are available to immediately ack an event. Queued events are
processed in mpt_free_request() where resources are freed. This
avoids a panic on resource shortage.
o Parse and print out RAID controller capabilities during driver probe.
o Define, allocate, and maintain RAID data structures for volumes,
hidden member physical disks and spare disks.
o Add dynamic sysctls for per-instance setting of the log level, array
resync rate, array member cache enable, and volume queue depth.
mpt_debug.c:
o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on
a particular log level to aid in tracking down driver issues.
o Add mpt_decode_value() which parses the bits in an integer
value based on a parsing table (mask, value, name string, tuples).
mpilib/*:
o Update mpi library header files to latest distribution from LSI.
Submitted by: gibbs
Approved by: re
place.
This moves the dependency on GCC's and other compiler's features into
the central sys/cdefs.h file, while the individual source files can
then refer to #ifdef __COMPILER_FEATURE_FOO where they by now used to
refer to #if __GNUC__ > 3.1415 && __BARC__ <= 42.
By now, GCC and ICC (the Intel compiler) have been actively tested on
IA32 platforms by netchild. Extension to other compilers is supposed
to be possible, of course.
Submitted by: netchild
Reviewed by: various developers on arch@, some time ago
Set the DMA SGL length correctly if the DMA request must be chained because
it is too large to fit in one SGL.
This should fix this driver for some Dell Precision systems.
RELENG_5 candidate.
PR: kern/66479
Submitted by: HITOSHI Osada <qfh02545@nifty.com>
to build the kernel. It doesn't affect the operation if gcc.
Most of the changes are just adding __INTEL_COMPILER to #ifdef's, as
icc v8 may define __GNUC__ some parts may look strange but are
necessary.
Additional changes:
- in_cksum.[ch]:
* use a generic C version instead of the assembly version in the !gcc
case (ASM code breaks with the optimizations icc does)
-> no bad checksums with an icc compiled kernel
Help from: andre, grehan, das
Stolen from: alpha version via ppc version
The entire checksum code should IMHO be replaced with the DragonFly
version (because it isn't guaranteed future revisions of gcc will
include similar optimizations) as in:
---snip---
Revision Changes Path
1.12 +1 -0 src/sys/conf/files.i386
1.4 +142 -558 src/sys/i386/i386/in_cksum.c
1.5 +33 -69 src/sys/i386/include/in_cksum.h
1.5 +2 -0 src/sys/netinet/igmp.c
1.6 +0 -1 src/sys/netinet/in.h
1.6 +2 -0 src/sys/netinet/ip_icmp.c
1.4 +3 -4 src/contrib/ipfilter/ip_compat.h
1.3 +1 -2 src/sbin/natd/icmp.c
1.4 +0 -1 src/sbin/natd/natd.c
1.48 +1 -0 src/sys/conf/files
1.2 +0 -1 src/sys/conf/files.amd64
1.13 +0 -1 src/sys/conf/files.i386
1.5 +0 -1 src/sys/conf/files.pc98
1.7 +1 -1 src/sys/contrib/ipfilter/netinet/fil.c
1.10 +2 -3 src/sys/contrib/ipfilter/netinet/ip_compat.h
1.10 +1 -1 src/sys/contrib/ipfilter/netinet/ip_fil.c
1.7 +1 -1 src/sys/dev/netif/txp/if_txp.c
1.7 +1 -1 src/sys/net/ip_mroute/ip_mroute.c
1.7 +1 -2 src/sys/net/ipfw/ip_fw2.c
1.6 +1 -2 src/sys/netinet/igmp.c
1.4 +158 -116 src/sys/netinet/in_cksum.c
1.6 +1 -1 src/sys/netinet/ip_gre.c
1.7 +1 -2 src/sys/netinet/ip_icmp.c
1.10 +1 -1 src/sys/netinet/ip_input.c
1.10 +1 -2 src/sys/netinet/ip_output.c
1.13 +1 -2 src/sys/netinet/tcp_input.c
1.9 +1 -2 src/sys/netinet/tcp_output.c
1.10 +1 -1 src/sys/netinet/tcp_subr.c
1.10 +1 -1 src/sys/netinet/tcp_syncache.c
1.9 +1 -2 src/sys/netinet/udp_usrreq.c
1.5 +1 -2 src/sys/netinet6/ipsec.c
1.5 +1 -2 src/sys/netproto/ipsec/ipsec.c
1.5 +1 -1 src/sys/netproto/ipsec/ipsec_input.c
1.4 +1 -2 src/sys/netproto/ipsec/ipsec_output.c
and finally remove
sys/i386/i386 in_cksum.c
sys/i386/include in_cksum.h
---snip---
- endian.h:
* DTRT in C++ mode
- quad.h:
* we don't use gcc v1 anymore, remove support for it
Suggested by: bde (long ago)
- assym.h:
* avoid zero-length arrays (remove dependency on a gcc specific
feature)
This change changes the contents of the object file, but as it's
only used to generate some values for a header, and the generator
knows how to handle this, there's no impact in the gcc case.
Explained by: bde
Submitted by: Marius Strobl <marius@alchemy.franken.de>
- aicasm.c:
* minor change to teach it about the way icc spells "-nostdinc"
Not approved by: gibbs (no reply to my mail)
- bump __FreeBSD_version (lang/icc needs to know about the changes)
Incarnations of this patch survive gcc compiles since a loooong time,
I use it on my desktop. An icc compiled kernel works since Nov. 2003
(exceptions: snd_* if used as modules), it survives a build of the
entire ports collection with icc.
Parts of this commit contains suggestions or submissions from
Marius Strobl <marius@alchemy.franken.de>.
Reviewed by: -arch
Submitted by: netchild
Add two new arguments to bus_dma_tag_create(): lockfunc and lockfuncarg.
Lockfunc allows a driver to provide a function for managing its locking
semantics while using busdma. At the moment, this is used for the
asynchronous busdma_swi and callback mechanism. Two lockfunc implementations
are provided: busdma_lock_mutex() performs standard mutex operations on the
mutex that is specified from lockfuncarg. dftl_lock() is a panic
implementation and is defaulted to when NULL, NULL are passed to
bus_dma_tag_create(). The only time that NULL, NULL should ever be used is
when the driver ensures that bus_dmamap_load() will not be deferred.
Drivers that do not provide their own locking can pass
busdma_lock_mutex,&Giant args in order to preserve the former behaviour.
sparc64 and powerpc do not provide real busdma_swi functions, so this is
largely a noop on those platforms. The busdma_swi on is64 is not properly
locked yet, so warnings will be emitted on this platform when busdma
callback deferrals happen.
If anyone gets panics or warnings from dflt_lock() being called, please
let me know right away.
Reviewed by: tmm, gibbs
Devices below may experience a change in geometry.
* Due to a bug, aic(4) never used extended geometry. Changes all drives
>1G to now use extended translation.
* sbp(4) drives exactly 1 GB in size now no longer use extended geometry.
* umass(4) drives exactly 1 GB in size now no longer use extended geometry.
For all other controllers in this commit, this should be a no-op.
Looked over by: scottl
BUS_DMASYNC_ definitions remain as before. The does not change the ABI,
and reverts the API to be a bit more compatible and flexible. This has
survived a full 'make universe'.
Approved by: re (bmah)
of 1024- Ultra4 256). Rename 'requests' tag to 'request_pool' for clarity.
Make sure we do correct xpt_freeze_simq/CAM_RELEASE_SIMQ if we run out
of chip resources.
MFC after: 6 days
In mpttimeout, call mpt_intr just on the offchance that we missed
an interrupt. We can check to see whether or not the command that
is timing out got completed.
When we *do* decide to timeout a command, set the command state to
REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2.
This allows us to catch a couple cases we've seen where the command
we timed out on in fact is ready to be completed by the firmware.
In any case, it's only after mpttimeout2 is called that we actually
take down the private state and free the request itself. CAM has
been notified in mpttimeout anyway. This whole area should be redone,
but that will take 105% of my available game time for this month.
Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK
and MPTLOCK_2_CAMLOCK locations.
Split mpt_notify into mpt_ctlop, which handles all reply completions
that have 0x800000000 or'd into the ContextID. This function can, in
fact, call mpt_event_notify_reply, which handles the traditional
async event notifications. While we're at it, put in the extremely
important (but currently untested) code that send back an Ack to
an Event Notification (if the Event Notification is marked with
AckRequired). Note that an Ack also generates another ctlop completion,
tra la.
Fix up mpt_done substantially to try and get how we plug into CAM
correctly done. Remove bogus CAM_RELEASE_SIMQ settings.
Do some cleanups in mpt_action that are related to speed negotiation
for Ultra4 cards. This is an area that is still quite fragile and
worrisome as config data being read back often doesn't make sense or
jibe with the documentation.
At any rate, after these changes were done, I was finally able to
get Lars Eggert's dual 320M disk system to stay up under load all
weekend- hopefully we're in good enough for now shape.
MFC after: 1 week
Define the CFG_DAGA_OFF offset as 128 bytes instead of 40- gives us
a more reasonable headroom.
When reading a config page, zero out the entire request area- not just
the length of the request. This is because we cleverly (cheezily) return
configuration data back into the allocated request area, so it's nice
to make sure we start with a clean area to write on.
MFC after: 1 week
a) we don't believe what the board tells us all the time (if the BIOS
hasn't run, port page 2 and port page 0 tend to be garbage)
b) add the missing code to set parameters for the SPI cards.
MFC after: 0 days
