Since the attribute container deletes from a klist while it's walking
it, it is vulnerable to the problem (and fix) here:

http://marc.theaimsgroup.com/?l=linux-scsi&m=112485448830217

The attached fixes this (but won't compile without the above).

It also fixes the logical reversal in the traversal loop which meant
that we were never actually traversing the loop to hit this bug in the
first place.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

One of the changes in the attribute_container code in the scsi-misc tree
was to add a lock to protect the list of devices per container.  This,
unfortunately, leads to potential scheduling while atomic problems if
there's a sleep in the function called by a trigger.

The correct solution is to use the kernel klist infrastructure instead
which allows lockless traversal of a list.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Here's the problem. Try to do this on 2.6.12:
- Kill udev and HAL
- Insert a CD-ROM into a SCSI or USB CD-ROM drive
- Run dd if=/dev/scd0
- cat /sys/block/sr0/size
- Eject the CD, insert a different one
- Run dd if=/dev/scd0
This is likely to do "access beyond the end of device", if you let it
- cat /sys/block/sr0/size
This shows the size of a previous CD, even though dd was supposed
to revalidate the device.
- Run dd if=/dev/scd0
The second run of dd works correctly!

The bug was introduced in 2.5.31, when Al fixes the recursive opens
in partitioning. Before, the code worked like this:
- Block layer called cdrom_open directly
- cdrom_open called sr_open
- sr_open called check_disk_change
- check_disk_change called sr_media_change
- sr_media_change did cd->needs_disk_change=1
- before returning sr_open tested cd->needs_disk_change
  and called get_sector_size.

In 2.6.12, the check_disk_change is called from cdrom_open only. Thus:
- Block layer calls sr_bd_open
- sr_bd_open calls cdrom_open
- cdrom_open calls sr_open
- sr_open tests cd->needs_disk_change, which wasn't set yet; returns
- cdrom_open calls check_disk_change
- check_disk_change calls sr_media_change
- sr_media_change does cd->needs_disk_change=1, but nobody cares

Acked by: Alexander Viro <aviro@redhat.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

This patch fixes a long term borkenness in
ibmvscsi where we were using the wrong timeout
field from the scsi command (and using the
wrong units.)  Now broken by the fact that the
scsi_cmnd timeout field is gone entirely.
This only worked before because all the SCSI
targets assumed that 0 was default.
Signed-off-by: Dave Boutcher <boutcher@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

C files should include the files with the prototypes for their global
functions.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

attribute_container_classdev_to_container is an exported function of the
attribute_container.c file.  However, there's no prototype for it.  Now
I actually want to use it, so add one.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

With the removal of the spinlocking around eh calls, we need to add a
little more locking back in, otherwise we do some naked list
manipulation.
Signed-off-by: Dave Boutcher <boutcher@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

This patch fixes the bad assumption of the aacraid driver with use_sg.
I used the 3w-xxxx driver fix as a guide for this.
Signed-off-by: Mark Haverkamp <markh@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

If your transport class sets the ATTRIBUTE_CONTAINER_NO_CLASSDEVS flag,
then its configure method never gets called.  This patch fixes that so
that the configure method is called with a NULL classdev.

Also remove a spurious inverted comma in the transport_class comments.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

remove lots of completely dead code from aiclib, there's not a lot left
and even what's left is rather useless.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

remove ahd_tailq and do sane pci probing.  ported over from aic7xxx.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

remove some dead cruft, as done already in aic7xxx
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

I recently tried to construct a totally generic transport class and
found there were certain features missing from the current abstract
transport class.  Most notable is that you have to hang the data on the
class_device but most of the API is framed in terms of the generic
device, not the class_device.

These changes are two fold

- Provide the class_device to all of the setup and configure APIs
- Provide and extra API to take the device and the attribute class and
  return the corresponding class_device
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

This patch is necessary if we begin exposing underlying physical disks
(which can attach to the SPI transport class) of the hardware RAID
cards, since we don't want any SPI parameters binding to the RAID
devices.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

From: Christoph Hellwig <hch@lst.de>

Multi-function cards need to inherit the PCI flags from the master PCI
device.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

From: "Martin J. Bligh" <mbligh@mbligh.org>

drivers/scsi/aic7xxx/aic7770.c: In function `aic7770_config':
drivers/scsi/aic7xxx/aic7770.c:129: warning: unused variable `l'
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

From: Andrew Morton <akpm@osdl.org>

drivers/scsi/scsi.c: In function `scsi_softirq':
drivers/scsi/scsi.c:814: warning: int format, long int arg (arg 4)
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Replace use of lpfc_put_lun with midlayer's int_to_scsilun

Remove driver's local definition of lpfc_put_lun (which converts an
int back to a 64-bit LUN) and replace it's use with the recently added
int_to_scsilun function provided by the midlayer.

Note: Embedding midlayer structure in our structure caused
need for more files to include midlayer headers.
Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Fix handling of the dev_loss and nodev timeouts.

Symptoms: when remote port disappears for a period of time longer then
either nodev_tmo or dev_loss_tmo, the lpfc driver worker thread will
stall removing that remote port.

Cause: removing remote port involves un-blocking and sync-ing
corresponding block device queue. But corresponding node in the lpfc
driver is still in the NPR(?node port recovery?) state and mid-layer
gets SCSI_MLQUEUE_HOST_BUSY as a return value when it is trying to call
queuecommand() with command for that node (AKA remote port)

Fix: Instead of returning SCSI_MLQUEUE_HOST_BUS from queuecommand() for
nodes in NPR states complete it with retry-able error code  DID_BUS_BUSY
Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Fix panic in lpfc_get_stats()

Symptoms: Panic on sysfs stats access

Cause: In lpfc_get_stats() we are writing to memory that we do not
own.

Fix: Fix our stats structure allocation. Embed phba->link_stats in
struct lpfc_hba and stop treating it like rogue structure.

Note: Embedding midlayer/transport structure in our structure caused
need for more files to include midlayer/transport headers.
Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Clear task management bits when preparing SCSI commands

In lpfc_scsi_prep_cmnd, clear the task management bits (fcpCntl2 member
in the fcp_cmd structure) when preparing regular SCSI commands.
Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Fix panic on lip and cable pull

Symptoms: Panic on lip or cable pull

Cause: Use after free of nlp in lpfc_nlp_remove()

Fix: Do not make FC transport calls after a node is removed. Transport
calls are disabled by ignoring the initial delete transition.
Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

IOCB BDE not getting fully initialized during reuse

Symptoms: Driver gets Status 3 and Reason 0x13 on IOCB completions.

Cause: The IOCB bpl.bdeSize and bdeFlags are not getting initialized on reuse.

Fix: Reinitialize these fields in prep_dma each time an IOCB is used.
Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

From: 	Steve Wilcox <spwilcox@att.com>

In order to properly report LUN's > 7, the DEC HSG80 definition in
scsi_devinfo.c needs to include BLIST_REPORTLUN2 rather than
BLIST_SPARSELUN.  I've tested this change with several HSG firmware
revisions and with both Emulex and Qlogic HBA's.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

There's a spurious (and illegal since it's marked __exit) call to
ahc_linux_exit() in ahc_linux_init() which causes a double list
deletion of the transport class; remove it.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

When run on a kernel that scans all LUNs, a certain crappy
scsi scanner reports the same LUN over and over..
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=155457

Aparently they were so shamed by this, they chose to remain
anonymous. Though it seems the blacklist code handles
anonymous vendors just fine.
Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Received from Mark Salyzyn

This patch adds the product ID for the ICP9067MA adapter.

The entries for the ICP9085LI, ICP5085BR, IBM8k & ASR4810SAS were
incorrect and would not initialize the adapters correctly.
Signed-off-by: Mark Haverkamp <markh@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

There are certain rogue devices (and the aic7xxx driver) that return
BUSY or QUEUE_FULL forever.  This code will apply a global timeout (of
the total number of retries times the per command timer) to a given
command.  If it is exceeded, the command is completed regardless of its
state.

The patch also removes the unused field in the command: timeout and
timeout_total.

This solves the problem of detecting an endless loop in the mid-layer
because of BUSY/QUEUE_FULL bouncing, but will not recover the device.
In the aic7xxx case, the driver can be recovered by sending a bus reset,
so possibly this should be tied into the error handler?
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

I have rediffed the patch against 2.6.13-rc5, done a couple of cosmetic
cleanups, and run some tests.  Brian King has acknowledged that it fixes the
problems he has seen. Seems mature enough for inclusion into 2.6.14 (or
later)?

Nate's explanation of the changes:

I've attached patches against 2.6.13rc2. These are basically identical
to my earlier patches, as I found that all issues I'd seen in earlier
kernels still existed in this kernel.

To summarize, the changes are: (more details in my original email)

- add a kref to the scsi_tape structure, and associate reference
counting stuff

- set sr_request->end_io = blk_end_sync_rq so we get notified when an IO
is rejected when the device goes away

- check rq_status when IOs complete, else we don't know that IOs
rejected for a dead device in fact did not complete

- change last_SRpnt so it's set before an async IO is issued (in case
st_sleep_done is bypassed)

- fix a bogus use of last_SRpnt in st_chk_result
Signed-off-by: Nate Dailey <nate.dailey@stratus.com>
Signed-off-by: Kai Makisara <kai.makisara@kolumbus.fi>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

The cmd->timeout field has been obsolete for a while now.  While looking
to remove it, I came across this use in the aacraid driver.  It looks
like you want to initialise the firmware with the current timeout of the
command (in seconds), so the value I think you should be using is
cmd->timeout_per_command.

Acked by: Mark Haverkamp <markh@osdl.org>
Acked by: Mark Salyzyn <mark_salyzyn@adaptec.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

without it you get this failure:

drivers/built-in.o(.text+0xdcccd): In function `ahd_linux_slave_configure':
drivers/scsi/aic7xxx/aic79xx_osm.c:636: undefined reference to `spi_dv_device'
drivers/built-in.o(.text+0xdd7b1): In function `ahd_send_async':
drivers/scsi/aic7xxx/aic79xx_osm.c:1652: undefined reference to `spi_display_xfer_agreement'
drivers/built-in.o(.init.text+0x7b4d): In function `ahd_linux_init':
drivers/scsi/aic7xxx/aic79xx_osm.c:2765: undefined reference to `spi_attach_transport'
drivers/built-in.o(.init.text+0x7c94):drivers/scsi/aic7xxx/aic79xx_osm.c:2774: undefined reference to `spi_release_transport'
drivers/built-in.o(.exit.text+0x72c): In function `ahd_linux_exit':
drivers/scsi/aic7xxx/aic79xx_osm.c:2783: undefined reference to `spi_release_transport'
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Received from Mark Salyzyn from Adaptec:

This patch adds support for the new raw io command. This new command
offers much larger io commands, is more friendly to the internal firmware
structure requiring less translation efforts by the firmware and offers
support for targets greater than 2TB (patch to support >2TB will
be sent in the future).
Signed-off-by: Mark Haverkamp <markh@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Received from Mark Salyzyn from Adaptec:

If the Adapter is quiet and does not produce an AIF event packets to be
picked up by the management applications for longer than the timeout
interval of two minutes, the cleanup code that deals with aging out
registrants could erroneously drop the registration. The timeout is
there to clean up should the management application die and fail to poll
for updated AIF event packets.

Moving the timer update from the ioctl code that delivers an AIF to the
polling registrant to the bottom of the ioctl means the timeout is reset
with any management application polling activity regardless if an AIF is
delivered or not removing the erroneous timeout cleanups.
Signed-off-by: Mark Haverkamp <markh@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Received from Mark Salyzyn from Adaptec:

This patch removes the duplicate code in the write_callback command
completion handler, and renames read_callback to io_callback. Optimized
the lba calculation into the debug print routine macro to optimize the
i/o code path.
Signed-off-by: Mark Haverkamp <markh@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Add in pci shutdown method so that the adapter shuts down correctly and
flushes its cache. Shutdown should also disable the adapter's interrupt
when shutdown (in particularly if the driver is rmmod'd) to prevent
spurious hardware activities.
Signed-off-by: Mark Haverkamp <markh@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Received from Mark Salyzyn from Adaptec.

Fixes a bug in check_revision.  It should return the driver version not
the firmware version.
Update driver version number.
Update driver version string.
Signed-off-by: Mark Haverkamp <markh@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Received from Mark Salyzyn from Adaptec:

If more than two commands are outstanding to the controller, there is no
need to notify the adapter via a PCI bus transaction of additional
commands added into the queue; it will get to them when it works through
the produce/consumer indexes.

This reduced the PCI traffic in the driver to submit a command to the
queue to near zero allowing a significant number of commands to be
turned around with no need to block for the PCI bridge to flush the
notify request to the adapter.

Interrupt mitigation has always been present in the driver; it was
turned off because of a bug that prevented one from realizing the
usefulness of the feature. This bug is fixed in this patch.
Signed-off-by: Mark Haverkamp <markh@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

I ran into this one sending bus resets across the hardware.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>