We were releasing the IRQ before removing the host, so commands could
still be coming in which would never be seen by the interrupt handler.
Just remove the host before releasing the IRQ to close this race.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Since ncr53c8xx_attach() calls scsi_host_put(), make ncr53c8xx_release()
call scsi_host_put() too, for symmetry. Both callers already expect
it to put the host for them, so that works out nicely. While the zalon
driver does 'use' the host pointer afterwards, it only compares it for
equality and doesn't dereference it, so that's safe.
While I'm at it, get rid of pointless checks for NULL, use shost_priv()
and change ncr53c8xx_release to return void.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
We were releasing the block devices before removing the host, so commands
could still be coming in which would cause a panic. Just remove the
host before releasing the block devices to close this race.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
There was a missing call to scsi_host_put() causing us to leak a scsi
host every time this module was unloaded.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
If kthread_run failed, we would fail to scan the host, and leak the
allocated async_scan_data. Since using a separate thread is just an
optimisation, do the scan synchronously if we fail to spawn a thread.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
The current code prints:
scsi 13:0:4:0: scsi: Device offlined - not ready after error recovery
which is repetitively redundant. This patch changes that message to:
scsi 6:0:6:0: Device offlined - not ready after error recovery
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
sg uses a scheme to reallocate a single contiguous array of all its
pointers for lookup and management. This didn't matter too much when sg
could only attach 256 nodes, but now the maximum has been bumped up to
32k we're starting to push the limits of the maximum allocatable
contiguous memory. The solution to this is to eliminate the static
array and do everything via idr, which this patch does.
Acked-by: Douglas Gilbert <dougg@torque.net>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
The OnStream SCSI Tape driver uses a semaphore as mutex. Use the mutex API
instead of the (binary) semaphore.
Signed-off-by: Matthias Kaehlcke <matthias.kaehlcke@gmail.com>
Reviewed-by: Satyam Sharma <satyam@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Willem Riede <wrlk@riede.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
* Remove IRQF_DISABLED, it is clearly wrong for this driver.
* Remove wasteful spin_lock_irqsave() in interrupt handler.
The lighter-weight spin_lock() is all that's needed.
* Annotate with FIXME where arcmsr_interrupt() is called
without any spinlock being acquired.
* Eliminate pointless cast from void pointer in arcmsr_do_interrupt()
[jejb: conflict resolution]
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Acked-by: Nick Cheng <nick.cheng@areca.com.tw>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Remove _interruptible, since receiving a signal while waiting on a
hardware condition will simply cause the driver to busy-wait.
Using msleep_interruptible() is rarely the right thing to do, when
waiting on a hardware condition to change.
Also, replace msleep with ssleep while doing this, where appropriate.
[jejb: fix up merge conflict]
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Acked-by: Nick Cheng <nick.cheng@areca.com.tw>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Do not adjust the iIDMA speed on ports which have a faster
link-speed than the HBA itself.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Since MSI-X vectors do not require a clearing "handshake" from
the system perspective, and the registered handler will not be
called more than once for one occurrence of receipt of a vector,
there is no requirement to flush the risc register write clearing
the interrupt condition in the risc. Also, since the msi-x
registered handlers are optimised for a particular vector, it is
preferable to handle the one vector received per invocation of
the handler.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Make several needlessly global functions static:
- qla2x00_mark_vp_devices_dead()
- qla24xx_configure_vp()
Remove unused function qla24xx_modify_vport().
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Drop usage of legacy to_qla_host() macro.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
This change reduces by as much as 16% the memory footprint for
each allocated sbr_t structure requested from the mempool.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Original code, incorrectly passed the address-of a pointer rather
than the pointer value itself.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Also remove legacy '/proc' name during OS_DEVICE_NAME
registration.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Where the DPC logic would get jammed into continuously
reloging-into a port.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
There's no need to reset the RISC prior to pausing.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Newer ISPs support a mechanism to read and write flash-memory via
the firmware LOAD/DUMP memory mailbox command routines. When
supported, utilizing these mechanisms significantly reduces
overall access times.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
One of the intents of the block prep function was to allow ULDs to use
it for preprocessing. The original SCSI model was to have a single prep
function and add a pointer indirect filter to build the necessary
commands. This patch reverses that, does away with the init_command
field of the scsi_driver structure and makes ULDs attach directly to the
prep function instead. The value is really that it allows us to begin
to separate the ULDs from the SCSI mid layer (as long as they don't use
any core functions---which is hard at the moment---a ULD doesn't even
need SCSI to bind).
Acked-by: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
If you build a multiplatform kernel for iSeries and pSeries, with
ibmvscsic support, the resulting client doesn't work on iSeries.
This fixes that, using the appropriate low-level operations
for the machine detected at runtime.
[jejb: fixed up rejections around the srp transport patch]
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Acked by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
So its ancient, its crap, but it kept showing up in my scans for stuff
that wanted fixing...
- Redo the proc code to be far cleaner
- Clean various return (0) type constructs
- Use cpu_relax()
The various waits ought to time out but thats another issue and probably
not worth solving.
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
In scsi module I've found some inconsistency between variable type
used in module_param_named and type passed to module_param_named as an
argument. Especially the inconsistency of `max_scsi_luns' parameter is
a bit serious because the description text says "last scsi LUN (should
be between 1 and 2^32-1)".
Signed-off-by: Masatake YAMATO <jet@gyve.org>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
- Convert ide-scsi to the new data accessors and cleanup
the !use_sg code paths.
In old code the driver was trying to translate narrow commands,
if received, to wide commands. This code path still assumed
scsi_cmnd->request_buffer is a linear char pointer.
This means that this driver was broken since 2.6.17.
As suggested by Christoph Hellwig I set
use_10_for_rw = 1; and use_10_for_ms = 1;
for this device and completely killed the translation.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Acked-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
A BUSY status returned on a write request results in a stale residual
being returned when the write ultimately successfully completes.
This can be reproduced as follows:
1) issue immediate mode rewind to scsi tape drive
2) issue write request
The tape drive returns busy. The low level driver detects underrun and
sets the residual into the scsi command. The low level driver responds
with (DID_OK << 16) | scsi_status. scsi_status is 8, hence
status_byte(result) == 4, i.e., BUSY.
scsi_softirq_done() calls scsi_decide_disposition() which returns
ADD_TO_MLQUEUE. scsi_softirq_done() then calls scsi_queue_insert()
which, on the way to resubmitting the request to the driver, calls
scsi_init_cmd_errh().
The attached patch modifies scsi_init_cmd_errh() to clear the resid
field. This prevents a "stale" residual from being returned when the
scsi command finally completes without a BUSY status.
Signed-off-by: Michael Reed <mdr@sgi.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
CC [M] drivers/scsi/arcmsr/arcmsr_hba.o
drivers/scsi/arcmsr/arcmsr_attr.c:186: warning: initialization from
incompatible pointer type
drivers/scsi/arcmsr/arcmsr_attr.c:196: warning: initialization from
incompatible pointer type
drivers/scsi/arcmsr/arcmsr_attr.c:206: warning: initialization from
incompatible pointer type
drivers/scsi/arcmsr/arcmsr_hba.c: In function 'arcmsr_alloc_ccb_pool':
drivers/scsi/arcmsr/arcmsr_hba.c:329: warning: assignment from
incompatible pointer type
drivers/scsi/arcmsr/arcmsr_hba.c: At top level:
drivers/scsi/arcmsr/arcmsr_hba.c:101: warning:
'arcmsr_pci_error_detected' declared 'static' but never defined
drivers/scsi/arcmsr/arcmsr_hba.c:102: warning: 'arcmsr_pci_slot_reset'
declared 'static' but never defined
The majority being incorrect casting or the fact that binary attributes
now take an additional argument.
Cc: Nick Cheng <nick.cheng@areca.com.tw>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Description:
** support ARC1200/1201/1202 SATA RAID adapter, which is named
ACB_ADAPTER_TYPE_B
** modify the arcmsr_pci_slot_reset function
** modify the arcmsr_pci_ers_disconnect_forepart function
** modify the arcmsr_pci_ers_need_reset_forepart function
Signed-off-by: Nick Cheng <nick.cheng@areca.com.tw>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
This removes the unused sysfs attribute overwriting logic for most of
the attributes, and plugs them into the driver core default attribute
creation.
Without this patch, at the time of the events for the SCSI LUN's, there
will be no sysfs files, because their creation is delayed until the sd
driver has spun up the disks, which might take several seconds. It is the
last WAIT_FOR_SYSFS rule in the default udev setup which can be removed
with this change.
Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Follow the example of some other drivers by defining DRV_NAME to be
"advansys". Prevents spelling mistakes.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
- Delete comments relating to the previous structure of the driver.
I have no intention of honouring them ;-)
- Reformat comments > 80 columns
- Remove now-obsolete comments from advansys_interrupt
- Change adv_get_sglist() from do {} while (1) to for (;;)
- Return void from AscInitQLinkVar()
- Take out a level of indentation in adv_get_sglist()
- Reduce indentation level of AscAsyncFix()
- Remove unused macros
- Refactor AscSendScsiQueue slightly
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
- Replace ASC_ASSERT() with BUG_ON().
In a few places, get rid of the assertion altogether -- the ensuing
crash will tell us all we need to know. Use BUG() where it fits better
than BUG_ON(). Also fix a fencepost error in advansys_proc_info().
- Replace DvcSleepMilliSecond with mdelay.
Despite its name using 'sleep', the implementation was a delay.
I've marked some places with XXX where we should probably be using
msleep instead. They need to be audited to be sure we can sleep in
that context.
- Replace DvcDelayMicroSecond with udelay.
- Replace DvcDelayNanoSecond with udelay too.
All callers were multiples of 1000.
- Remove DvcEnterCritical and DvcLeaveCritical.
These functions are no-ops, and as the comments said, the spinlock
protects the critical sections.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
AscExeScsiQueue() has one caller, and it passes the address of a
variable; this cannot ever be NULL. This is the only place that ever
sets ASCQ_ERR_SCSIQ_NULL_PTR, so delete that error code too, as well as
several other unused ASCQ_ERR codes.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
The SCSI midlayer won't send commands greater than ->max_cmd_len.
So the checks on length in asc_build_req and adv_build_req are obsolete
and can be deleted, but also we have to set the max_cmd_len in
advansys_board_found(). Also move the length definitions together, and
write a helpful comment.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Move as much as possible outside the critical section in queuecommand, eg:
- Set the scsi_done field before acquiring the lock
- Call asc_scsi_done after dropping the lock
Also remove a comment suggesting we should enable interrupts (now we do)
and do some minor reformatting for readability.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
The `active' list was used in the reset handler -- but the midlayer
guarantees that list is empty by the time the reset handler is invoked.
It was also checked in the interrupt routines to be sure that this command
belonged to this board, but we don't need to check that either. We can
then delete the asc_prt_target_stats() function as it will never print
anything, along with asc_enqueue(), asc_dequeue_list(), asc_rmqueue(),
asc_scsi_done_list(), struct asc_queue, ASC_QUEUE_EMPTY, ASC_TID_ALL,
ASC_FRONT, ASC_BACK, and all the REQ* macros. Also remove this item from
the todo list.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
- Move the guts of asc_scsi_done_list() into a new function, asc_scsi_done.
- Call asc_scsi_done() in asc_isr_callback() and adv_isr_callback(). The
comment was wrong; scsi_done cannot enable interrupts.
- All other places which queued an scp on the done list are error paths
for queuecommand, and so we can just call asc_scsi_done() in queuecommand
if we receive an error.
- We no longer need to keep a list of done requests in advansys_interrupt
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
The midlayer guarantees it won't call ->queuecommand for a host which
is handling a reset condition.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
If the adapter is busy, return the request to the midlayer rather than
queueing it in the driver. asc_execute_queue() and asc_dequeue() become
unused, and we don't need to print out stats on the waiting queue any more.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Move the error reporting into AscInitGetConfig, AdvInitGetConfig and
AscInitSetConfig.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
